chore(release): v3.22.0

Co-authored-by: Codex <noreply@openai.com>

README.md

@@ -21,7 +21,7 @@
 
 ## 快速预览
 
-无需部署，打开 [快速预览服务](https://codex-proxy-rs.ainz.cc) 即可体验管理端的系统概览、账号分组、代理管理与用量统计
+无需部署即可体验管理端的系统概览、账号分组、代理管理与用量统计：
 
 | 登录信息 | 默认值 |
 | --- | --- |
@@ -53,8 +53,7 @@ curl -fsSL https://raw.githubusercontent.com/zyycn/codex-proxy-rs/main/deploy/in
 [安装脚本](deploy/install.sh) 默认安装到当前目录下的 `codex-proxy-rs/`，下载同一正式 Release 的部署文件，
 自动生成密码、设置目录权限并启动服务
 
-完成后会显示访问地址和管理员密码，请保存密码，再按下方步骤
-[添加账号与客户端密钥](#添加账号与客户端密钥)
+完成后保存显示的访问地址和管理员密码，再[登录管理端](#登录管理端)并[添加账号与客户端密钥](#添加账号与客户端密钥)
 
 可在执行 `bash install.sh` 时传入环境变量：
 

backend/Cargo.lock

@@ -303,7 +303,7 @@ dependencies = [
  "lru",
  "percent-encoding",
  "regex-lite",
- "sha2 0.11.0",
+ "sha2",
  "tracing",
  "url",
 ]
@@ -325,7 +325,7 @@ dependencies = [
  "hmac",
  "http",
  "percent-encoding",
- "sha2 0.11.0",
+ "sha2",
  "time",
  "tracing",
 ]
@@ -358,7 +358,7 @@ dependencies = [
  "md-5",
  "pin-project-lite",
  "sha1 0.11.0",
- "sha2 0.11.0",
+ "sha2",
  "tracing",
 ]
 
@@ -877,7 +877,7 @@ dependencies = [
  "gateway-plugin-sdk",
  "hex",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "tar",
  "tempfile",
  "thiserror",
@@ -1688,7 +1688,7 @@ dependencies = [
  "semver",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "subtle",
  "syn 3.0.6",
  "tempfile",
@@ -1719,7 +1719,7 @@ dependencies = [
  "secrecy",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "thiserror",
  "tokio",
  "tokio-tungstenite 0.29.0",
@@ -1745,12 +1745,11 @@ dependencies = [
  "futures",
  "futures-timer",
  "http",
- "http-body",
  "http-body-util",
  "semver",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "thiserror",
  "tracing",
  "url",
@@ -1790,7 +1789,7 @@ dependencies = [
  "serde_ignored",
  "serde_json",
  "serde_path_to_error",
- "sha2 0.11.0",
+ "sha2",
  "tar",
  "tempfile",
  "thiserror",
@@ -1831,7 +1830,7 @@ dependencies = [
  "semver",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "sqlx",
  "tar",
  "tempfile",
@@ -1859,6 +1858,7 @@ dependencies = [
 name = "gateway-protocol"
 version = "0.0.0"
 dependencies = [
+ "http",
  "httpdate",
  "memchr",
  "serde",
@@ -1887,7 +1887,7 @@ dependencies = [
  "secrecy",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "sqlx",
  "tempfile",
  "thiserror",
@@ -2525,6 +2525,15 @@ version = "0.2.186"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "68ab91017fe16c622486840e4c83c9a37afeff978bd239b5293d61ece587de66"
 
+[[package]]
+name = "libredox"
+version = "0.1.25"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "61ff90caf6077a803a240f62fdbe88645a890bbca49ef8174c3cb0404362171d"
+dependencies = [
+ "libc",
+]
+
 [[package]]
 name = "libsqlite3-sys"
 version = "0.37.0"
@@ -2842,6 +2851,24 @@ dependencies = [
  "libc",
 ]
 
+[[package]]
+name = "objc2-core-foundation"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "2a180dd8642fa45cdb7dd721cd4c11b1cadd4929ce112ebd8b9f5803cc79d536"
+dependencies = [
+ "bitflags",
+]
+
+[[package]]
+name = "objc2-system-configuration"
+version = "0.3.2"
+source = "registry+https://github.com/rust-lang/crates.io-index"
+checksum = "7216bd11cbda54ccabcab84d523dc93b858ec75ecfb3a7d89513fa22464da396"
+dependencies = [
+ "objc2-core-foundation",
+]
+
 [[package]]
 name = "once_cell"
 version = "1.21.4"
@@ -2930,7 +2957,7 @@ dependencies = [
  "elliptic-curve",
  "primefield",
  "primeorder",
- "sha2 0.11.0",
+ "sha2",
 ]
 
 [[package]]
@@ -3271,7 +3298,7 @@ dependencies = [
  "semver",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "subtle",
  "tar",
  "tempfile",
@@ -3313,7 +3340,7 @@ dependencies = [
  "semver",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "thiserror",
  "tokio",
  "tracing",
@@ -4024,17 +4051,6 @@ version = "1.0.1"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "bbfa15b3dddfee50a0fff136974b3e1bde555604ba463834a7eb7deb6417705d"
 
-[[package]]
-name = "sha2"
-version = "0.10.9"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "a7507d819769d01a365ab707794a4084392c824f54a7a6a7862f8c3d0892b283"
-dependencies = [
- "cfg-if",
- "cpufeatures 0.2.17",
- "digest 0.10.7",
-]
-
 [[package]]
 name = "sha2"
 version = "0.11.0"
@@ -4155,8 +4171,7 @@ dependencies = [
 [[package]]
 name = "sqlx"
 version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "378620ccc25c62c89d8be1c819e76a88d59bdcc3304733330788948e619bfd71"
+source = "git+https://github.com/transact-rs/sqlx?rev=f1e94ec34dd5f11ad6488a3a8ac7988439a3f757#f1e94ec34dd5f11ad6488a3a8ac7988439a3f757"
 dependencies = [
  "sqlx-core",
  "sqlx-macros",
@@ -4168,8 +4183,7 @@ dependencies = [
 [[package]]
 name = "sqlx-core"
 version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "05b44e85bf579a8eeb4ceaa77a3a523baf2bf0e9bac7e40f405d537b5d2d5ccb"
+source = "git+https://github.com/transact-rs/sqlx?rev=f1e94ec34dd5f11ad6488a3a8ac7988439a3f757#f1e94ec34dd5f11ad6488a3a8ac7988439a3f757"
 dependencies = [
  "base64 0.22.1",
  "bytes",
@@ -4192,7 +4206,7 @@ dependencies = [
  "rustls",
  "serde",
  "serde_json",
- "sha2 0.10.9",
+ "sha2",
  "smallvec",
  "thiserror",
  "tokio",
@@ -4206,8 +4220,7 @@ dependencies = [
 [[package]]
 name = "sqlx-macros"
 version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "bd2b84f2bc39a5705ef27ec785a11c934a41bbd4a24941e257927cddc26b60bf"
+source = "git+https://github.com/transact-rs/sqlx?rev=f1e94ec34dd5f11ad6488a3a8ac7988439a3f757#f1e94ec34dd5f11ad6488a3a8ac7988439a3f757"
 dependencies = [
  "proc-macro2",
  "quote",
@@ -4219,8 +4232,7 @@ dependencies = [
 [[package]]
 name = "sqlx-macros-core"
 version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "fb8d96de5fdc85a5c4ec813432b523ec637e80ba98f046555f75f7908ddac7c3"
+source = "git+https://github.com/transact-rs/sqlx?rev=f1e94ec34dd5f11ad6488a3a8ac7988439a3f757#f1e94ec34dd5f11ad6488a3a8ac7988439a3f757"
 dependencies = [
  "cfg-if",
  "dotenvy",
@@ -4231,7 +4243,7 @@ dependencies = [
  "quote",
  "serde",
  "serde_json",
- "sha2 0.10.9",
+ "sha2",
  "sqlx-core",
  "sqlx-mysql",
  "sqlx-postgres",
@@ -4245,8 +4257,7 @@ dependencies = [
 [[package]]
 name = "sqlx-mysql"
 version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "90b8020fe17c5f2c245bfa2505d7ef59c5604839527c740266ad2214acebea27"
+source = "git+https://github.com/transact-rs/sqlx?rev=f1e94ec34dd5f11ad6488a3a8ac7988439a3f757#f1e94ec34dd5f11ad6488a3a8ac7988439a3f757"
 dependencies = [
  "bitflags",
  "byteorder",
@@ -4263,7 +4274,7 @@ dependencies = [
  "percent-encoding",
  "serde",
  "sha1 0.11.0",
- "sha2 0.11.0",
+ "sha2",
  "sqlx-core",
  "thiserror",
  "tracing",
@@ -4273,8 +4284,7 @@ dependencies = [
 [[package]]
 name = "sqlx-postgres"
 version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "87a2bdd6e83f6b3ea525ca9fee568030508b58355a43d0b2c1674d5f79dcd65e"
+source = "git+https://github.com/transact-rs/sqlx?rev=f1e94ec34dd5f11ad6488a3a8ac7988439a3f757#f1e94ec34dd5f11ad6488a3a8ac7988439a3f757"
 dependencies = [
  "atoi",
  "base64 0.22.1",
@@ -4297,7 +4307,7 @@ dependencies = [
  "rand 0.10.2",
  "serde",
  "serde_json",
- "sha2 0.11.0",
+ "sha2",
  "smallvec",
  "sqlx-core",
  "stringprep",
@@ -4310,8 +4320,7 @@ dependencies = [
 [[package]]
 name = "sqlx-sqlite"
 version = "0.9.0"
-source = "registry+https://github.com/rust-lang/crates.io-index"
-checksum = "488e99c397a62007e4229aec669a179816339afc6d2620ca6fa420dbee2e982c"
+source = "git+https://github.com/transact-rs/sqlx?rev=f1e94ec34dd5f11ad6488a3a8ac7988439a3f757#f1e94ec34dd5f11ad6488a3a8ac7988439a3f757"
 dependencies = [
  "atoi",
  "chrono",
@@ -4494,7 +4503,7 @@ source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "32497e9a4c7b38532efcdebeef879707aa9f794296a4f0244f6f69e9bc8574bd"
 dependencies = [
  "fastrand",
- "getrandom 0.4.3",
+ "getrandom 0.3.4",
  "once_cell",
  "rustix",
  "windows-sys 0.61.2",
@@ -5183,6 +5192,12 @@ name = "whoami"
 version = "2.1.2"
 source = "registry+https://github.com/rust-lang/crates.io-index"
 checksum = "998767ef88740d1f5b0682a9c53c24431453923962269c2db68ee43788c5a40d"
+dependencies = [
+ "libc",
+ "libredox",
+ "objc2-system-configuration",
+ "web-sys",
+]
 
 [[package]]
 name = "winapi"

backend/Cargo.toml

@@ -59,3 +59,5 @@ panic = "abort"
 [patch.crates-io]
 # Upstream SOCKS IPv6 fix; remove once a release includes hyperium/hyper-util#302.
 hyper-util = { git = "https://github.com/hyperium/hyper-util", rev = "191b1acf0a6b18e807e4abf5a6d4a7c47979ecd6" }
+# 固定上游 BEGIN 取消回滚修复，避免未结束事务复用；见 transact-rs/sqlx#4394
+sqlx = { git = "https://github.com/transact-rs/sqlx", rev = "f1e94ec34dd5f11ad6488a3a8ac7988439a3f757" }

backend/apps/gateway/src/bootstrap/command.rs

@@ -93,7 +93,7 @@ async fn execute(
         Some(plugin_runtime.clone()),
         Some(plugin_runtime.observer_registry()),
         Some(plugin_runtime.policy_registry()),
-        Some(plugin_runtime.middleware_registry()),
+        Some(plugin_runtime.execution_registry()),
     );
     let management = ManagementPorts::new(&store, providers.admin.clone(), core.snapshot_control());
     or_shutdown!(plugin_runtime, management.bind(&plugin_runtime));
@@ -110,7 +110,7 @@ async fn execute(
         providers.admin,
         std::sync::Arc::new(gateway_host::pricing::ModelsDevPricing),
     );
-    let mut services = gateway_admin::service::Registry::new(service_middleware(
+    let mut services = gateway_admin::public_service::Registry::new(service_middleware(
         core.snapshots(),
         &plugin_runtime,
     ));

backend/apps/gateway/src/bootstrap/plugins.rs

@@ -72,11 +72,11 @@ impl ManagementPorts {
 pub(super) fn service_middleware(
     snapshots: RuntimeSnapshotHandle,
     runtime: &PluginRuntime,
-) -> gateway_admin::service::PlanSource {
-    let middleware = runtime.middleware_registry();
+) -> gateway_admin::public_service::PlanSource {
+    let middleware = runtime.execution_registry();
     Arc::new(move || {
         let snapshot = snapshots.snapshot_for_diagnostics()?;
-        middleware.resolve(snapshot.extensions()?)
+        middleware.middleware(snapshot.extensions()?)
     })
 }
 

backend/apps/gateway/src/bootstrap/server.rs

@@ -28,7 +28,7 @@ pub async fn run() -> Result<(), BootstrapError> {
         Some(plugin_runtime.clone()),
         Some(plugin_runtime.observer_registry()),
         Some(plugin_runtime.policy_registry()),
-        Some(plugin_runtime.middleware_registry()),
+        Some(plugin_runtime.execution_registry()),
         Some(plugin_runtime.frontend_authentication_registry()),
     );
     let management = ManagementPorts::new(&store, providers.admin.clone(), core.snapshot_control());
@@ -122,14 +122,14 @@ pub async fn run() -> Result<(), BootstrapError> {
     )
     .with_middleware({
         let snapshots = core.snapshots();
-        let middleware = plugin_runtime.middleware_registry();
+        let middleware = plugin_runtime.execution_registry();
         move |snapshot| {
             if let Some(snapshot) = snapshot {
-                return middleware.resolve(snapshot.extensions()?);
+                return middleware.middleware(snapshot.extensions()?);
             }
             // 数据面未就绪时，管理与诊断路由仍可使用发布候选中的插件
             let diagnostic = snapshots.snapshot_for_diagnostics()?;
-            middleware.resolve(diagnostic.extensions()?)
+            middleware.middleware(diagnostic.extensions()?)
         }
     });
     let http_dispatcher = api.dispatcher();

backend/apps/gateway/tests/architecture.rs

@@ -79,18 +79,25 @@ fn core_value_owners_do_not_depend_on_execution_or_routing() {
             path if path.starts_with("policy/") => {
                 Some(&["account", "identity", "policy", "validation"])
             }
-            "settings/values.rs" | "settings/compiled.rs" => {
-                Some(&["account", "concurrency", "identity", "metering", "policy"])
-            }
+            path if path.starts_with("settings/") => Some(&[
+                "account",
+                "concurrency",
+                "identity",
+                "metering",
+                "policy",
+                "settings",
+            ]),
             path if path.starts_with("account/") => Some(&["account", "identity", "validation"]),
             _ => None,
         };
         let Some(allowed) = allowed else { continue };
         let source = fs::read_to_string(root.join(&relative)).expect("read core source");
         let syntax = syn::parse_file(&source).expect("parse core source");
-        let mut references = CrateReferences::default();
+        let mut references = CrateReferences::in_module(&relative);
         references.visit_file(&syntax);
-        for dependency in references.0 {
+        // 相对路径中的 self 指向当前 owner，不是跨 owner 依赖
+        references.owners.remove(&references.module_path[0]);
+        for dependency in references.owners {
             assert!(
                 allowed.contains(&dependency.as_str()),
                 "{} must not depend on crate::{dependency}",
@@ -100,49 +107,155 @@ fn core_value_owners_do_not_depend_on_execution_or_routing() {
     }
 }
 
+#[test]
+fn core_owner_paths_form_a_dag() {
+    let root = backend_root().join("crates/gateway-core/src");
+    let mut graph = BTreeMap::<String, BTreeSet<String>>::new();
+    for relative in super::rust_files(&root) {
+        if relative == Path::new("lib.rs") {
+            continue;
+        }
+        let mut references = CrateReferences::in_module(&relative);
+        let owner = references.module_path[0].clone();
+        let source = fs::read_to_string(root.join(&relative)).expect("read core source");
+        references.visit_file(&syn::parse_file(&source).expect("parse core source"));
+        references.owners.remove(&owner);
+        graph.entry(owner).or_default().extend(references.owners);
+    }
+    for owner in graph.keys() {
+        let mut pending = graph[owner].iter().cloned().collect::<Vec<_>>();
+        let mut visited = BTreeSet::new();
+        while let Some(dependency) = pending.pop() {
+            assert_ne!(
+                &dependency, owner,
+                "Core owner dependency cycle reaches {owner}"
+            );
+            if visited.insert(dependency.clone())
+                && let Some(next) = graph.get(&dependency)
+            {
+                pending.extend(next.iter().cloned());
+            }
+        }
+    }
+}
+
+#[test]
+fn core_reference_scan_resolves_groups_reexports_and_relative_paths() {
+    let syntax = syn::parse_file(
+        r#"
+        pub use crate::{engine::AttemptContext, settings::SettingsValues as Values};
+        use super::super::runtime as publication;
+        use self::compiled::Value;
+        // crate::ignored::Comment
+        const EXAMPLE: &str = "crate::ignored::Literal";
+        "#,
+    )
+    .expect("parse reference fixture");
+    let mut references = CrateReferences::in_module(Path::new("routing/snapshot.rs"));
+    references.visit_file(&syntax);
+    assert_eq!(
+        references.owners,
+        BTreeSet::from([
+            "engine".to_owned(),
+            "settings".to_owned(),
+            "runtime".to_owned(),
+            "routing".to_owned(),
+        ])
+    );
+}
+
 #[derive(Default)]
-struct CrateReferences(BTreeSet<String>);
+struct CrateReferences {
+    owners: BTreeSet<String>,
+    module_path: Vec<String>,
+}
 
 impl<'ast> Visit<'ast> for CrateReferences {
     fn visit_path(&mut self, path: &'ast syn::Path) {
-        let mut segments = path.segments.iter();
-        if segments
-            .next()
-            .is_some_and(|segment| segment.ident == "crate")
-            && let Some(owner) = segments.next()
-        {
-            self.0.insert(owner.ident.to_string());
-        }
+        self.record(
+            &path
+                .segments
+                .iter()
+                .map(|part| part.ident.to_string())
+                .collect::<Vec<_>>(),
+        );
         syn::visit::visit_path(self, path);
     }
 
     fn visit_item_use(&mut self, item: &'ast syn::ItemUse) {
-        self.use_tree(&item.tree, false);
+        self.use_tree(&item.tree, &mut Vec::new());
     }
 }
 
 impl CrateReferences {
-    fn use_tree(&mut self, tree: &syn::UseTree, crate_root: bool) {
+    fn in_module(relative: &Path) -> Self {
+        let mut module = relative
+            .with_extension("")
+            .components()
+            .map(|part| {
+                part.as_os_str()
+                    .to_str()
+                    .expect("UTF-8 module path")
+                    .to_owned()
+            })
+            .collect::<Vec<_>>();
+        if module.last().is_some_and(|part| part == "mod") {
+            module.pop();
+        }
+        Self {
+            owners: BTreeSet::new(),
+            module_path: module,
+        }
+    }
+
+    fn record(&mut self, path: &[String]) {
+        let Some(first) = path.first() else { return };
+        let owner = if first == "crate" {
+            path.get(1).cloned()
+        } else if first == "self" || first == "super" {
+            let mut module = self.module_path.clone();
+            let mut rest = path;
+            while rest
+                .first()
+                .is_some_and(|part| part == "super" || part == "self")
+            {
+                if rest[0] == "super" {
+                    module.pop();
+                }
+                rest = &rest[1..];
+            }
+            module.first().or_else(|| rest.first()).cloned()
+        } else {
+            None
+        };
+        if let Some(owner) = owner {
+            self.owners.insert(owner);
+        }
+    }
+
+    fn use_tree(&mut self, tree: &syn::UseTree, prefix: &mut Vec<String>) {
         match tree {
-            syn::UseTree::Path(path) if crate_root => {
-                self.0.insert(path.ident.to_string());
+            syn::UseTree::Path(path) => {
+                prefix.push(path.ident.to_string());
+                self.use_tree(&path.tree, prefix);
+                prefix.pop();
             }
-            syn::UseTree::Name(name) if crate_root => {
-                self.0.insert(name.ident.to_string());
+            syn::UseTree::Name(name) => {
+                prefix.push(name.ident.to_string());
+                self.record(prefix);
+                prefix.pop();
             }
-            syn::UseTree::Rename(rename) if crate_root => {
-                self.0.insert(rename.ident.to_string());
+            syn::UseTree::Rename(rename) => {
+                prefix.push(rename.ident.to_string());
+                self.record(prefix);
+                prefix.pop();
             }
-            syn::UseTree::Path(path) => self.use_tree(&path.tree, path.ident == "crate"),
             syn::UseTree::Group(group) => {
                 for item in &group.items {
-                    self.use_tree(item, crate_root);
+                    self.use_tree(item, prefix);
                 }
             }
-            syn::UseTree::Glob(_) if crate_root => {
-                self.0.insert("*".to_owned());
-            }
-            _ => {}
+            syn::UseTree::Glob(_) => self.record(prefix),
         }
     }
 }
@@ -221,7 +334,7 @@ const ADAPTER_PUBLIC_MODULES: &[(&str, &[&str])] = &[
         "crates/providers/openai",
         &["config", "credential", "transport"],
     ),
-    ("crates/providers/xai", &["credential", "transport"]),
+    ("crates/providers/xai", &["transport"]),
 ];
 
 /// 不对应单一生产模块、而是校验 crate/workspace 整体契约的根级测试场景

backend/apps/gateway/tests/bootstrap/config.rs

@@ -76,11 +76,6 @@ fn config_loader_should_reject_missing_runtime_data_dir() {
     assert!(parse_config(&config).is_err());
 }
 
-#[test]
-fn config_loader_should_inject_connection_passwords_into_urls() {
-    parse_config(&valid_config()).expect("Store validates password injection into both URLs");
-}
-
 #[test]
 fn config_loader_should_apply_only_explicit_topology_overrides() {
     let invalid = valid_config()

backend/apps/plugin-cli/README.md

@@ -28,11 +28,8 @@ cpr-plugin package \
 
 资源映射相对于作者清单所在目录解析。输出包括完整插件 ID、版本和目标 triple 命名的 `.tar.gz` 及 `.sha256`；归档根目录直接包含 `plugin.json`、可执行文件和已声明资源，不包含开发依赖
 
-作者清单可省略普通贡献项的 `id`、`version` 和固定阶段；`middleware` 必须显式声明版本 `4`，
-并从 `http`、`websocket`、`service`、`request`、`attempt` 中选择挂载阶段；`upstream_adapter` 必须显式声明版本 `2`。
-CLI 与 SDK 的 `Manifest::from_author_slice`、`PluginBuilder::from_json` 共用规范化入口；归档中的
-`plugin.json` 写入补全后的贡献项字段，不保留另一份作者声明。
-完整规则见 [SDK 清单](../../crates/gateway-plugin/sdk/docs/manifest.md)
+作者清单的省略规则、能力版本和挂载阶段见 [SDK 扩展项声明](../../crates/gateway-plugin/sdk/docs/manifest.md#扩展项简写)。
+CLI 与 SDK 的 `Manifest::from_author_slice`、`PluginBuilder::from_json` 共用规范化入口，归档只保留补全字段后的 `plugin.json`
 
 支持 `x86_64-unknown-linux-gnu`、`aarch64-unknown-linux-gnu` 和 `aarch64-apple-darwin`。
 完整开发示例在独立仓库 `codex-proxy-plugins` 的 `examples/workbench`，可通过[子模块](../../../docs/development.md#源码联调)检出；本工具不负责源码构建、上传或启用插件

backend/crates/gateway-admin/src/backup/task.rs

@@ -281,7 +281,6 @@ impl BackupTask {
                 .await?;
             }
             None => {
-                let _ = self.dump.cleanup_staging(&record.id).await;
                 self.fail_task(record, code::PG_DUMP_FAILED, "导出进程中断，暂存归档不完整")
                     .await?;
             }
@@ -307,7 +306,8 @@ impl BackupTask {
             && record.sha256.as_deref() == Some(remote.sha256.as_str())
         {
             let now = Utc::now();
-            self.repository
+            if self
+                .repository
                 .transition_status(
                     &record.id,
                     status_transition(BackupStatus::Uploading, BackupStatus::Completed)?,
@@ -315,7 +315,11 @@ impl BackupTask {
                     now,
                 )
                 .await
-                .map_err(repo_error)?;
+                .map_err(repo_error)?
+                .is_some()
+            {
+                self.cleanup_staging(&record.id).await;
+            }
             return Ok(());
         }
         match self.dump.inspect_staging(&record.id).await? {
@@ -455,9 +459,7 @@ impl BackupTask {
             now,
             cancellation,
         )
-        .await?;
-        let _ = self.dump.cleanup_staging(&record.id).await;
-        Ok(())
+        .await
     }
 
     /// 上传 + 远端校验，内部完成终态迁移
@@ -518,6 +520,7 @@ impl BackupTask {
                         size_bytes,
                         "备份完成"
                     );
+                    self.cleanup_staging(&record.id).await;
                 }
             }
             Ok(Some(remote)) => {
@@ -596,10 +599,16 @@ impl BackupTask {
             )
             .await
             .map_err(repo_error)?;
-        let _ = self.dump.cleanup_staging(&record.id).await;
+        self.cleanup_staging(&record.id).await;
         Ok(())
     }
 
+    async fn cleanup_staging(&self, backup_id: &str) {
+        if let Err(error) = self.dump.cleanup_staging(backup_id).await {
+            warn!(backup_id, error = %error, "清理备份暂存失败");
+        }
+    }
+
     /// 执行一小批到期保留清理：先处理 `expires_at` 已到期的记录，再按天数/份数清理计划备份
     async fn run_retention_batch(&self, now: DateTime<Utc>) -> Result<(), WorkerTaskError> {
         let settings = self.repository.load_settings().await.map_err(repo_error)?;

backend/crates/gateway-admin/src/lib.rs

@@ -20,7 +20,7 @@ pub mod backup;
 pub mod freeze_recovery;
 pub mod model;
 pub mod ports;
-pub mod service;
+pub mod public_service;
 mod use_case;
 pub use use_case::plugins::{PluginDistributionPorts, PluginManagementService, PluginsService};
 
@@ -205,7 +205,7 @@ pub enum AdminConfigError {
 #[derive(Clone)]
 pub struct AdminServices {
     timezone: gateway_core::time::DeploymentTimeZone,
-    public_services: Arc<service::Registry>,
+    public_services: Arc<public_service::Registry>,
     plugins: Arc<PluginsService>,
     plugin_management: Arc<PluginManagementService>,
     proxies: Arc<dyn ProxiesService>,
@@ -219,7 +219,8 @@ pub struct AdminServices {
     settings: Arc<dyn SettingsService>,
     system: Arc<dyn SystemService>,
     credentials: Arc<CredentialsService>,
-    plugin_accounts: Arc<dyn PluginAccountAccess>,
+    // Runtime 只持有 Weak，这个强引用保障账号端口与 Admin 服务同寿命
+    _plugin_accounts: Arc<dyn PluginAccountAccess>,
     backups: Arc<dyn BackupService>,
     import_tasks: Arc<dyn ImportTasksService>,
 }
@@ -230,7 +231,7 @@ impl AdminServices {
         self.timezone
     }
 
-    pub fn public_services(&self) -> Arc<service::Registry> {
+    pub fn public_services(&self) -> Arc<public_service::Registry> {
         self.public_services.clone()
     }
 
@@ -254,12 +255,6 @@ impl AdminServices {
         self.key_usage.as_ref()
     }
 
-    /// 取得账号服务的共享句柄；后台编排（冻结恢复 worker）需要持有 Arc
-    #[must_use]
-    pub fn accounts_handle(&self) -> Arc<dyn AccountsService> {
-        Arc::clone(&self.accounts)
-    }
-
     #[must_use]
     pub fn proxies(&self) -> &dyn ProxiesService {
         self.proxies.as_ref()
@@ -309,12 +304,6 @@ impl AdminServices {
         self.credentials.as_ref()
     }
 
-    /// Runtime 只持有该窄端口的 Weak；AdminBundle 保持实际生命周期
-    #[must_use]
-    pub fn plugin_accounts_handle(&self) -> Arc<dyn PluginAccountAccess> {
-        Arc::clone(&self.plugin_accounts)
-    }
-
     #[must_use]
     pub fn backups(&self) -> &dyn BackupService {
         self.backups.as_ref()
@@ -342,7 +331,7 @@ impl AdminBundle {
 /// 组合根提供给控制面的运行能力；与配置和存储端口分别传入
 pub struct AdminRuntimePorts {
     pub timezone: gateway_core::time::DeploymentTimeZone,
-    pub service_middleware: service::PlanSource,
+    pub service_middleware: public_service::PlanSource,
     pub plugin_preparation: Arc<dyn ports::plugins::PluginPreparation>,
     pub plugin_management: Arc<dyn ports::plugin_management::PluginManagement>,
     pub published_snapshot: gateway_core::runtime::RuntimeSnapshotHandle,
@@ -477,7 +466,7 @@ async fn initialize_inner(
         registry.clone(),
         pricing_source,
     );
-    let mut public_services = service::Registry::new(service_middleware);
+    let mut public_services = public_service::Registry::new(service_middleware);
     public_services.register_settings(&settings)?;
     let services = AdminServices {
         timezone,
@@ -526,7 +515,7 @@ async fn initialize_inner(
         settings,
         system,
         credentials,
-        plugin_accounts,
+        _plugin_accounts: plugin_accounts,
         import_tasks,
         backups,
     };

backend/crates/gateway-admin/src/model/account_groups.rs

@@ -126,6 +126,16 @@ pub struct AccountGroupPage {
     pub page_size: u16,
 }
 
+/// 分组选择器使用的轻量分页结果
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct AccountGroupOptionsPage {
+    pub config_revision: Revision,
+    pub items: Vec<AccountGroupRef>,
+    pub total: u64,
+    pub page: u32,
+    pub page_size: u16,
+}
+
 /// 创建账号分组
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct CreateAccountGroup {

backend/crates/gateway-admin/src/model/observability.rs

@@ -177,6 +177,32 @@ pub struct UsageQuery {
     pub page_size: ObservabilityPageSize,
 }
 
+/// Admin 确定概览的数量、趋势粒度及最近请求筛选，Store 只执行该选择
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct DashboardQuery {
+    pub range: TimeRange,
+    pub granularity: Granularity,
+    pub account_limit: u16,
+    pub recent_request_limit: u16,
+    pub recent_request_filter: UsageFilter,
+}
+
+impl DashboardQuery {
+    #[must_use]
+    pub fn new(range: TimeRange) -> Self {
+        Self {
+            range,
+            granularity: Granularity::for_range(range),
+            account_limit: 4,
+            recent_request_limit: 10,
+            recent_request_filter: UsageFilter {
+                outcome: Some(RequestOutcome::Succeeded),
+                ..UsageFilter::default()
+            },
+        }
+    }
+}
+
 /// 运维错误过滤条件
 #[derive(Debug, Clone, Default, PartialEq, Eq)]
 pub struct OpsErrorFilter {
@@ -477,6 +503,30 @@ pub enum Granularity {
     Day,
 }
 
+impl Granularity {
+    /// 管理趋势按范围选择粒度，费用校验与请求指标必须使用同一策略
+    #[must_use]
+    pub fn for_range(range: TimeRange) -> Self {
+        let seconds = range.end.signed_duration_since(range.start).num_seconds();
+        if seconds <= 2 * 24 * 60 * 60 {
+            Self::FifteenMinutes
+        } else if seconds <= 31 * 24 * 60 * 60 {
+            Self::Hour
+        } else {
+            Self::Day
+        }
+    }
+
+    #[must_use]
+    pub const fn seconds(self) -> i64 {
+        match self {
+            Self::FifteenMinutes => 15 * 60,
+            Self::Hour => 60 * 60,
+            Self::Day => 24 * 60 * 60,
+        }
+    }
+}
+
 /// 一段时间桶内的请求指标
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct RequestMetricPoint {
@@ -589,13 +639,14 @@ pub struct DashboardObservation {
     pub range: TimeRange,
     pub totals: DashboardTotals,
     pub provider_accounts: AccountPoolMetrics,
+    pub runtime_slots: Option<DashboardRuntimeSlots>,
     pub trend: Vec<RequestMetricPoint>,
     pub account_usage: Vec<DashboardAccountUsage>,
     pub recent_requests: Vec<UsageListRecord>,
 }
 
 /// 使用记录表格的窄读模型
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct UsageListRecord {
     pub client_api_key_name: Option<String>,
     pub id: String,
@@ -636,6 +687,13 @@ pub struct UsageListRecord {
     pub first_text_ms: Option<u64>,
     pub first_token_ms: Option<u64>,
     pub provider_processing_ms: Option<u64>,
+    pub upstream_response_ms: Option<u64>,
+    pub upstream_api_overhead_ms: Option<f64>,
+    pub upstream_engine_ms: Option<f64>,
+    pub upstream_engine_iapi_ttft_ms: Option<f64>,
+    pub upstream_engine_service_ttft_ms: Option<f64>,
+    pub upstream_engine_iapi_tbt_ms: Option<f64>,
+    pub upstream_engine_service_tbt_ms: Option<f64>,
     pub latency_ms: Option<u64>,
     pub admission_decision_ms: Option<u64>,
     pub account_selection_wait_ms: Option<u64>,
@@ -651,7 +709,7 @@ pub struct UsageListRecord {
 }
 
 /// 一次完整模型请求的公共观测记录
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct UsageRecord {
     pub id: String,
     pub client_api_key_ref: String,
@@ -710,6 +768,13 @@ pub struct UsageRecord {
     pub first_text_ms: Option<u64>,
     pub first_token_ms: Option<u64>,
     pub provider_processing_ms: Option<u64>,
+    pub upstream_response_ms: Option<u64>,
+    pub upstream_api_overhead_ms: Option<f64>,
+    pub upstream_engine_ms: Option<f64>,
+    pub upstream_engine_iapi_ttft_ms: Option<f64>,
+    pub upstream_engine_service_ttft_ms: Option<f64>,
+    pub upstream_engine_iapi_tbt_ms: Option<f64>,
+    pub upstream_engine_service_tbt_ms: Option<f64>,
     pub latency_ms: Option<u64>,
     pub admission_decision_ms: Option<u64>,
     pub account_selection_wait_ms: Option<u64>,
@@ -730,7 +795,7 @@ pub struct UsageRecord {
 }
 
 /// 用量分页结果
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct UsagePage {
     pub items: Vec<UsageListRecord>,
     pub current_page: u32,
@@ -776,7 +841,7 @@ pub struct UsageAttempt {
 }
 
 /// 一条请求及其全部尝试
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct UsageDetail {
     pub trace: Option<serde_json::Value>,
     pub related_requests: Vec<serde_json::Value>,

backend/crates/gateway-admin/src/ports/plugin_management.rs

@@ -1,7 +1,7 @@
 //! 插件管理页面、资源访问与授权回调的运行时端口
 
 use async_trait::async_trait;
-use gateway_core::runtime::extensions::ExtensionSetReference;
+use gateway_core::routing::extensions::ExtensionSetReference;
 
 use crate::model::{
     AdminError,

backend/crates/gateway-admin/src/ports/plugins.rs

@@ -26,7 +26,7 @@ use crate::model::{
         PluginCompatibilityRequirements, PluginSource,
     },
 };
-use gateway_core::runtime::extensions::ExtensionSetReference;
+use gateway_core::routing::extensions::ExtensionSetReference;
 
 /// Admin 准备完整候选并保活到提交后的发布结束；进程状态不能代替持久启用状态
 #[async_trait]
@@ -175,12 +175,13 @@ pub trait PluginDistribution: Send + Sync {
 /// Runtime 只解释插件包格式；安装事务与来源选择归 Admin
 #[async_trait]
 pub trait PluginPackageInspector: Send + Sync {
-    /// 可解析但版本范围未经宿主承诺的诊断，不决定启动资格
+    /// 完整校验包并返回版本范围诊断；校验失败与可加载但未经宿主承诺的提醒分开
     async fn compatibility_warning(
         &self,
-        _archive: Arc<[u8]>,
-        _expected_sha256: String,
+        archive: Arc<[u8]>,
+        expected_sha256: String,
     ) -> Result<Option<String>, AdminError> {
+        self.inspect(archive, Some(expected_sha256)).await?;
         Ok(None)
     }
 

backend/crates/gateway-admin/src/ports/store.rs

@@ -12,8 +12,9 @@ use super::backup::BackupStorePorts;
 use crate::model::{
     MutationContext, Revision,
     account_groups::{
-        AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation, AccountGroupPage,
-        DeleteAccountGroup, NewAccountGroup, SetAccountGroupEnabled, UpdateAccountGroup,
+        AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation,
+        AccountGroupOptionsPage, AccountGroupPage, DeleteAccountGroup, NewAccountGroup,
+        SetAccountGroupEnabled, UpdateAccountGroup,
     },
     accounts::{
         AccountListQuery, AccountPage, AccountPageItem, AccountRuntimeSnapshot,
@@ -27,9 +28,9 @@ use crate::model::{
         UpdateClientKey, UpdateClientKeyBudgetLimits,
     },
     observability::{
-        DashboardObservation, DashboardRuntimeSlots, DiagnosticDimension, DiagnosticsObservation,
-        OpsErrorPage, OpsErrorQuery, RequestMetricPoint, TimeRange, UsageCalculatedBillingFact,
-        UsageDetail, UsageFilter, UsageOverview, UsagePage, UsageQuery,
+        DashboardObservation, DashboardQuery, DiagnosticDimension, DiagnosticsObservation,
+        Granularity, OpsErrorPage, OpsErrorQuery, RequestMetricPoint, TimeRange,
+        UsageCalculatedBillingFact, UsageDetail, UsageFilter, UsageOverview, UsagePage, UsageQuery,
     },
     provider_credentials::{
         AuthorizationCommit, CredentialDetails, CredentialImportCommit, CredentialImportResult,
@@ -385,6 +386,11 @@ pub trait AccountGroupStore: Send + Sync {
         query: AccountGroupListQuery,
     ) -> AdminStoreResult<AccountGroupPage>;
 
+    async fn list_account_group_options(
+        &self,
+        query: AccountGroupListQuery,
+    ) -> AdminStoreResult<AccountGroupOptionsPage>;
+
     async fn load_account_group_members(
         &self,
         group_ids: &[gateway_core::routing::AccountGroupId],
@@ -425,27 +431,21 @@ pub trait ObservabilityStore: Send + Sync {
     /// 返回历史统计区间和指定观测时刻下的实时账号状态
     async fn dashboard_summary(
         &self,
-        range: TimeRange,
+        query: DashboardQuery,
         observed_at: DateTime<Utc>,
     ) -> AdminStoreResult<DashboardObservation>;
 
-    /// 返回 Dashboard 可选的实时槽位事实
-    ///
-    /// 该状态来自可丢失的运行时存储；无实现或运行时存储不可用时返回 `None`，不影响
-    /// 持久观测数据的读取
-    async fn dashboard_runtime_slots(
+    async fn dashboard_trend(
         &self,
-        _observed_at: DateTime<Utc>,
-    ) -> AdminStoreResult<Option<DashboardRuntimeSlots>> {
-        Ok(None)
-    }
-
-    async fn dashboard_trend(&self, range: TimeRange) -> AdminStoreResult<Vec<RequestMetricPoint>>;
+        range: TimeRange,
+        granularity: Granularity,
+    ) -> AdminStoreResult<Vec<RequestMetricPoint>>;
 
     async fn usage_trend(
         &self,
         range: TimeRange,
         filter: UsageFilter,
+        granularity: Granularity,
     ) -> AdminStoreResult<Vec<RequestMetricPoint>>;
 
     /// 流式返回可由 Provider 重新校验的已计算费用事实，不保证顺序
@@ -454,6 +454,7 @@ pub trait ObservabilityStore: Send + Sync {
         &self,
         range: TimeRange,
         filter: UsageFilter,
+        granularity: Granularity,
     ) -> UsageCalculatedBillingStream<'_>;
 
     async fn list_usage_records(&self, query: UsageQuery) -> AdminStoreResult<UsagePage>;
@@ -471,6 +472,7 @@ pub trait ObservabilityStore: Send + Sync {
         range: TimeRange,
         filter: UsageFilter,
         dimension: DiagnosticDimension,
+        limit: u16,
     ) -> AdminStoreResult<DiagnosticsObservation>;
 
     async fn list_ops_errors(&self, query: OpsErrorQuery) -> AdminStoreResult<OpsErrorPage>;

backend/crates/gateway-admin/src/public_service/mod.rs

@@ -8,11 +8,11 @@ use std::{future::Future, sync::Arc};
 
 use futures::future::BoxFuture;
 use gateway_core::{
-    engine::middleware::FrozenMiddlewarePlan,
-    middleware::{
-        compose,
+    engine::middleware::{
+        FrozenMiddlewarePlan,
         service::{Context, Error, Value},
     },
+    middleware::compose,
 };
 
 use crate::model::{AdminError, AdminErrorKind};

backend/crates/gateway-admin/src/public_service/registry.rs

@@ -2,7 +2,7 @@
 
 use crate::model::AdminError;
 use futures::future::BoxFuture;
-use gateway_core::middleware::service::{Error, Value};
+use gateway_core::engine::middleware::service::{Error, Value};
 use serde::{Serialize, de::DeserializeOwned};
 use std::{collections::BTreeMap, sync::Arc};
 

backend/crates/gateway-admin/src/use_case/account_groups.rs

@@ -19,9 +19,9 @@ use crate::{
         AdminError, MutationContext,
         account_groups::{
             AccountGroupAccountSummary, AccountGroupCapacity, AccountGroupListQuery,
-            AccountGroupMemberFact, AccountGroupMutation, AccountGroupPage, AccountGroupRecord,
-            CreateAccountGroup, DeleteAccountGroup, NewAccountGroup, SetAccountGroupEnabled,
-            UpdateAccountGroup,
+            AccountGroupMemberFact, AccountGroupMutation, AccountGroupOptionsPage,
+            AccountGroupPage, AccountGroupRecord, CreateAccountGroup, DeleteAccountGroup,
+            NewAccountGroup, SetAccountGroupEnabled, UpdateAccountGroup,
         },
     },
     ports::store::{AccountGroupStore, AccountRuntimeStore},
@@ -33,6 +33,10 @@ use super::{map_store_error, publish_committed};
 #[async_trait]
 pub trait AccountGroupService: Send + Sync {
     async fn list(&self, query: AccountGroupListQuery) -> Result<AccountGroupPage, AdminError>;
+    async fn list_options(
+        &self,
+        query: AccountGroupListQuery,
+    ) -> Result<AccountGroupOptionsPage, AdminError>;
     async fn create(
         &self,
         context: &MutationContext,
@@ -130,6 +134,16 @@ impl AccountGroupService for DefaultAccountGroupService {
         Ok(page)
     }
 
+    async fn list_options(
+        &self,
+        query: AccountGroupListQuery,
+    ) -> Result<AccountGroupOptionsPage, AdminError> {
+        self.store
+            .list_account_group_options(query)
+            .await
+            .map_err(|error| map_store_error(error, "account group options"))
+    }
+
     async fn create(
         &self,
         context: &MutationContext,

backend/crates/gateway-admin/src/use_case/credential_mutation.rs

@@ -0,0 +1,371 @@
+//! 凭据变更共享的绑定校验、事务结算、失效与提交后观察
+
+use crate::{
+    model::{
+        AdminError, MutationContext,
+        accounts::DeleteAccounts,
+        provider_credentials::{
+            AuthorizationMutationTarget, CredentialDeletion, CredentialDeletionResult,
+            CredentialDetails, CredentialMutationResult, CredentialRotationCommit,
+            PendingAuthorizationMutation, PreparedAuthorizationCommit,
+            PreparedAuthorizationCredential, PreparedCredentialImport, PreparedCredentialRotation,
+            ProviderQuotaRequest, StartAuthorization,
+        },
+    },
+    ports::{provider::ProviderAdmin, store::AccountStore},
+};
+use gateway_core::{account::ProviderAccountId, routing::ProviderKind, runtime::SnapshotControl};
+
+use super::{map_store_error, publish_committed};
+
+pub(super) async fn publish_credentials_and_observe_quota(
+    provider: &std::sync::Arc<dyn crate::ports::provider::ProviderAdmin>,
+    snapshot: &dyn SnapshotControl,
+    revision: crate::model::Revision,
+    account_ids: &[ProviderAccountId],
+    request_id: &str,
+) -> Result<(), AdminError> {
+    provider.account_facts_changed(account_ids).await;
+    publish_committed(snapshot, revision).await?;
+
+    // 凭据已经提交；额度只是可重建的观察，不能拖住管理请求或回滚提交结果
+    let provider = provider.clone();
+    let account_ids = account_ids.to_vec();
+    let request_id = request_id.to_owned();
+    tokio::spawn(async move {
+        for account_id in account_ids {
+            if let Err(error) = provider
+                .quota(ProviderQuotaRequest {
+                    account_id: account_id.clone(),
+                    refresh: true,
+                    rolling_usage: None,
+                })
+                .await
+            {
+                tracing::warn!(
+                    request_id,
+                    provider = %provider.provider_kind().as_str(),
+                    account_id = %account_id.as_str(),
+                    quota_error = ?error.kind(),
+                    "initial quota observation failed"
+                );
+            }
+        }
+    });
+    Ok(())
+}
+
+pub(super) async fn required_credential(
+    accounts: &dyn AccountStore,
+    provider_kind: &ProviderKind,
+    account_id: &ProviderAccountId,
+    resource: &'static str,
+) -> Result<CredentialDetails, AdminError> {
+    accounts
+        .credential_details(provider_kind, account_id)
+        .await
+        .map_err(|error| map_store_error(error, resource))?
+        .ok_or_else(|| AdminError::not_found("Provider 凭据不存在"))
+}
+
+pub(super) async fn required_plugin_credential(
+    accounts: &dyn AccountStore,
+    account_id: &ProviderAccountId,
+    resource: &'static str,
+) -> Result<CredentialDetails, AdminError> {
+    accounts
+        .credential_details_by_id(account_id)
+        .await
+        .map_err(|error| map_store_error(error, resource))?
+        .ok_or_else(|| AdminError::not_found("Provider 凭据不存在"))
+}
+
+pub(super) async fn import_proxy_binding(
+    proxies: &dyn crate::ports::proxy::ProxyStore,
+    id: Option<&str>,
+) -> Result<Option<crate::ports::proxy::ProxyImportReservation>, AdminError> {
+    let Some(id) = id else { return Ok(None) };
+    let reservation = proxies
+        .reserve_import(id)
+        .await
+        .map_err(|error| map_store_error(error, "proxy"))?;
+    Ok(Some(reservation))
+}
+
+pub(super) async fn pending_authorization(
+    accounts: &dyn AccountStore,
+    proxies: &dyn crate::ports::proxy::ProxyStore,
+    provider_kind: &ProviderKind,
+    command: &StartAuthorization,
+    resource: &'static str,
+) -> Result<PendingAuthorizationMutation, AdminError> {
+    let proxy;
+    let mut proxy_id = None;
+    let target = match &command.reauthorization {
+        Some(account_id) => {
+            proxy = required_credential(accounts, provider_kind, account_id, resource)
+                .await?
+                .credential
+                .outbound_proxy;
+            AuthorizationMutationTarget::Reauthorize {
+                account_id: account_id.clone(),
+            }
+        }
+        None => {
+            use crate::model::proxies::AccountProxySelection;
+            proxy = match &command.outbound_proxy {
+                Some(AccountProxySelection::Url(value)) => Some(value.clone()),
+                Some(AccountProxySelection::Saved(id)) => {
+                    let record = proxies
+                        .get(id)
+                        .await
+                        .map_err(|error| map_store_error(error, "proxy"))?;
+                    proxy_id = Some(record.id);
+                    Some(record.proxy)
+                }
+                Some(AccountProxySelection::Direct) | None => None,
+            };
+            AuthorizationMutationTarget::Create {
+                name: command.name.clone(),
+            }
+        }
+    };
+    Ok(PendingAuthorizationMutation::new(
+        provider_kind.clone(),
+        target,
+        crate::model::provider_credentials::AuthorizationOwnerBinding::from_context(
+            &command.context,
+        ),
+    )
+    .with_outbound_proxy(proxy)
+    .with_outbound_proxy_id(proxy_id))
+}
+
+pub(super) fn validate_prepared_import(
+    provider_kind: &ProviderKind,
+    prepared: &PreparedCredentialImport,
+    _resource: &'static str,
+) -> Result<(), AdminError> {
+    if prepared.provider_kind != *provider_kind
+        || prepared
+            .credentials
+            .iter()
+            .any(|credential| credential.provider_kind != *provider_kind)
+    {
+        return Err(AdminError::conflict("Provider 准备结果与请求范围不一致"));
+    }
+    Ok(())
+}
+
+pub(super) fn validate_prepared_rotation(
+    account: &crate::model::accounts::AccountRecord,
+    prepared: &PreparedCredentialRotation,
+    _resource: &'static str,
+) -> Result<(), AdminError> {
+    validate_prepared_rotation_facts(account, prepared.facts())
+}
+
+pub(super) fn validate_prepared_rotation_facts(
+    account: &crate::model::accounts::AccountRecord,
+    facts: &crate::model::provider_credentials::PreparedCredentialRotationFacts,
+) -> Result<(), AdminError> {
+    if facts.account_id.as_str() != account.id.as_str()
+        || facts.provider_kind != account.provider_kind
+    {
+        return Err(AdminError::conflict("Provider 准备结果与当前凭据不一致"));
+    }
+    Ok(())
+}
+
+pub(super) async fn validate_authorization_commit(
+    provider_kind: &ProviderKind,
+    context: &MutationContext,
+    prepared: PreparedAuthorizationCommit,
+    resource: &'static str,
+) -> Result<PreparedAuthorizationCommit, AdminError> {
+    if prepared.pending.provider_kind() != provider_kind
+        || !prepared.pending.owner_binding().matches_context(context)
+    {
+        let validation_error = AdminError::conflict("OAuth 授权绑定已失效，请重新发起授权");
+        if let Err(error) = prepared.abort().await {
+            tracing::warn!(
+                resource,
+                settlement_error = %error,
+                "OAuth authorization claim release failed after preparation validation"
+            );
+            return Err(error);
+        }
+        return Err(validation_error);
+    }
+    let matches_target = match (prepared.pending.target(), &prepared.credential) {
+        (
+            AuthorizationMutationTarget::Create { .. },
+            PreparedAuthorizationCredential::Create(credential),
+        ) => credential.provider_kind == *provider_kind,
+        (
+            AuthorizationMutationTarget::Reauthorize { account_id },
+            PreparedAuthorizationCredential::Reauthorize(credential),
+        ) => {
+            let facts = credential.facts();
+            facts.provider_kind == *provider_kind && facts.account_id == *account_id
+        }
+        _ => false,
+    };
+    if !matches_target {
+        let validation_error = AdminError::conflict("OAuth 凭据与待提交目标不一致");
+        if let Err(error) = prepared.abort().await {
+            tracing::warn!(
+                resource,
+                settlement_error = %error,
+                "OAuth authorization claim release failed after preparation validation"
+            );
+            return Err(error);
+        }
+        return Err(validation_error);
+    }
+    Ok(prepared)
+}
+
+pub(super) async fn commit_authorization(
+    accounts: &dyn AccountStore,
+    prepared: PreparedAuthorizationCommit,
+    key: crate::model::provider_credentials::AuthorizationReceiptKey,
+    settings: Option<crate::model::accounts::AccountImportSettings>,
+    context: &MutationContext,
+    resource: &'static str,
+) -> Result<crate::model::provider_credentials::AuthorizationCommitResult, AdminError> {
+    if settings.is_some()
+        && matches!(
+            &prepared.credential,
+            PreparedAuthorizationCredential::Reauthorize(_)
+        )
+    {
+        prepared.abort().await?;
+        return Err(AdminError::invalid("重新授权不能修改账号设置"));
+    }
+    let crate::model::provider_credentials::AuthorizationCommitSettlement {
+        command,
+        credential_guard,
+        authorization_guard,
+    } = prepared.into_commit(settings, key);
+    match accounts.commit_authorization(command, context).await {
+        Ok(result) => {
+            if let Some(guard) = credential_guard
+                && result.newly_committed
+            {
+                guard.finish();
+            }
+            if let Some(guard) = authorization_guard
+                && let Err(error) = guard.commit().await
+            {
+                // 账号与回执已经原子提交，Redis 清理失败不能把确定的成功改写为失败
+                tracing::warn!(resource, settlement_error = %error, "authorization committed but pending cleanup failed");
+            }
+            Ok(result)
+        }
+        Err(error) => {
+            drop(credential_guard);
+            if let Some(guard) = authorization_guard
+                && let Err(settlement_error) = guard.abort().await
+            {
+                tracing::warn!(
+                    resource,
+                    store_error = %error,
+                    settlement_error = %settlement_error,
+                    "OAuth authorization claim release failed after Store commit failure"
+                );
+                return Err(settlement_error);
+            }
+            Err(map_store_error(error, resource))
+        }
+    }
+}
+
+pub(super) async fn commit_credential_rotation(
+    accounts: &dyn AccountStore,
+    prepared: PreparedCredentialRotation,
+    settings: Option<crate::model::accounts::UpdateAccount>,
+    context: &MutationContext,
+    resource: &'static str,
+) -> Result<CredentialMutationResult, AdminError> {
+    let (facts, guard) = prepared.into_parts();
+    match accounts
+        .commit_credential_rotation(
+            CredentialRotationCommit {
+                prepared: facts,
+                settings,
+            },
+            context,
+        )
+        .await
+    {
+        Ok(result) => {
+            guard.finish();
+            Ok(result)
+        }
+        Err(error) => {
+            drop(guard);
+            Err(map_store_error(error, resource))
+        }
+    }
+}
+
+pub(super) async fn commit_credential_refresh(
+    accounts: &dyn AccountStore,
+    prepared: PreparedCredentialRotation,
+    context: &MutationContext,
+    resource: &'static str,
+) -> Result<CredentialMutationResult, AdminError> {
+    let (facts, guard) = prepared.into_parts();
+    match accounts
+        .commit_credential_refresh(
+            CredentialRotationCommit {
+                prepared: facts,
+                settings: None,
+            },
+            context,
+        )
+        .await
+    {
+        Ok(result) => {
+            guard.finish();
+            Ok(result)
+        }
+        Err(error) => {
+            drop(guard);
+            Err(map_store_error(error, resource))
+        }
+    }
+}
+
+pub(super) async fn delete_credentials(
+    accounts: &dyn AccountStore,
+    provider: &dyn ProviderAdmin,
+    command: CredentialDeletion,
+    resource: &'static str,
+) -> Result<CredentialDeletionResult, AdminError> {
+    for account_id in &command.account_ids {
+        required_credential(accounts, provider.provider_kind(), account_id, resource).await?;
+    }
+    let account_ids = command.account_ids;
+    let revision = accounts
+        .delete_accounts(
+            DeleteAccounts {
+                account_ids: account_ids
+                    .iter()
+                    .map(|account_id| account_id.as_str().to_owned())
+                    .collect(),
+            },
+            &command.context,
+        )
+        .await
+        .map_err(|error| map_store_error(error, resource))?;
+    for account_id in &account_ids {
+        provider.account_unavailable(account_id).await;
+    }
+    provider.account_facts_changed(&account_ids).await;
+    Ok(CredentialDeletionResult {
+        config_revision: revision,
+        account_ids,
+    })
+}

backend/crates/gateway-admin/src/use_case/error.rs

@@ -0,0 +1,67 @@
+//! 管理端口错误到用例错误的统一分类与来源保留
+
+use crate::model::{AdminError, AdminErrorKind};
+use crate::ports::store::{AdminStoreError, AdminStoreErrorKind};
+
+pub(super) fn map_store_error(error: AdminStoreError, resource: &'static str) -> AdminError {
+    let kind = match error.kind() {
+        AdminStoreErrorKind::Invalid => AdminErrorKind::Invalid,
+        AdminStoreErrorKind::NotFound => AdminErrorKind::NotFound,
+        AdminStoreErrorKind::StaleRevision
+        | AdminStoreErrorKind::DuplicateName
+        | AdminStoreErrorKind::Conflict => AdminErrorKind::Conflict,
+        AdminStoreErrorKind::Unavailable => AdminErrorKind::Unavailable,
+    };
+    let message = match kind {
+        AdminErrorKind::Invalid => "请求参数不合法",
+        AdminErrorKind::NotFound => "请求的资源不存在",
+        AdminErrorKind::Conflict => "当前资源状态冲突，请刷新后重试",
+        AdminErrorKind::Unavailable => "依赖服务暂不可用",
+        _ => "服务内部错误",
+    };
+    AdminError::new(kind, message).with_source(OperationFailure {
+        operation: resource,
+        source: error.into(),
+    })
+}
+
+pub(super) fn map_provider_error(
+    error: crate::ports::provider::ProviderAdminError,
+    resource: &'static str,
+) -> AdminError {
+    use crate::ports::provider::ProviderAdminErrorKind;
+
+    let kind = match error.kind() {
+        ProviderAdminErrorKind::Invalid => AdminErrorKind::Invalid,
+        ProviderAdminErrorKind::Unsupported => AdminErrorKind::Invalid,
+        ProviderAdminErrorKind::NotFound => AdminErrorKind::NotFound,
+        ProviderAdminErrorKind::Conflict => AdminErrorKind::Conflict,
+        ProviderAdminErrorKind::Ambiguous => AdminErrorKind::UpstreamResultUnknown,
+        ProviderAdminErrorKind::Unavailable => AdminErrorKind::Unavailable,
+        ProviderAdminErrorKind::CredentialRefreshRequired => AdminErrorKind::Unavailable,
+        ProviderAdminErrorKind::BadGateway => AdminErrorKind::BadGateway,
+        ProviderAdminErrorKind::Internal => AdminErrorKind::Internal,
+    };
+    let message = match kind {
+        AdminErrorKind::Invalid => "Provider 请求不合法",
+        AdminErrorKind::NotFound => "Provider 资源不存在",
+        AdminErrorKind::Conflict => "Provider 资源状态冲突，请刷新后重试",
+        AdminErrorKind::BadGateway => "上游服务请求失败",
+        AdminErrorKind::UpstreamResultUnknown => "上游执行结果未知，请刷新状态后再决定是否重试",
+        AdminErrorKind::Unavailable => "Provider 服务暂不可用",
+        AdminErrorKind::Internal => "服务内部错误",
+        _ => "Provider 操作失败",
+    };
+    AdminError::new(kind, error.public_message().unwrap_or(message)).with_source(OperationFailure {
+        operation: resource,
+        source: error.into(),
+    })
+}
+
+// 用例的操作名补充端口资源上下文，底层分类与原始原因仍由来源持有
+#[derive(Debug, thiserror::Error)]
+#[error("admin operation failed: {operation}")]
+struct OperationFailure {
+    operation: &'static str,
+    source: gateway_core::error::ErrorSource,
+}

backend/crates/gateway-admin/src/use_case/key_usage.rs

@@ -14,7 +14,9 @@ use crate::{
             KeyUsageOverview, KeyUsageQuery, KeyUsageRecordKind, KeyUsageRecords,
             KeyUsageRecordsQuery,
         },
-        observability::{OpsErrorFilter, OpsErrorQuery, TimeRange, UsageFilter, UsageQuery},
+        observability::{
+            Granularity, OpsErrorFilter, OpsErrorQuery, TimeRange, UsageFilter, UsageQuery,
+        },
         system::SystemVersion,
     },
     ports::store::{ClientKeyStore, ObservabilityStore},
@@ -192,9 +194,13 @@ impl KeyUsageService for DefaultKeyUsageService {
         };
         let (overview, trend, health_points) = futures::try_join!(
             self.observations.usage_summary(query.range, filter.clone()),
-            self.observations.usage_trend(query.range, filter),
             self.observations
-                .usage_trend(today, usage_filter(&id, None)),
+                .usage_trend(query.range, filter, Granularity::for_range(query.range)),
+            self.observations.usage_trend(
+                today,
+                usage_filter(&id, None),
+                Granularity::for_range(today)
+            ),
         )
         .map_err(|error| map_store_error(error, "key usage overview"))?;
         Ok(Some(KeyUsageOverview {

backend/crates/gateway-admin/src/use_case/mod.rs

@@ -19,450 +19,16 @@ pub mod proxies;
 pub mod settings;
 pub mod system;
 
-use crate::{
-    model::{
-        AdminError, AdminErrorKind, MutationContext,
-        accounts::DeleteAccounts,
-        provider_credentials::{
-            AuthorizationMutationTarget, CredentialDeletion, CredentialDeletionResult,
-            CredentialDetails, CredentialMutationResult, CredentialRotationCommit,
-            PendingAuthorizationMutation, PreparedAuthorizationCommit,
-            PreparedAuthorizationCredential, PreparedCredentialImport, PreparedCredentialRotation,
-            ProviderQuotaRequest, StartAuthorization,
-        },
-    },
-    ports::{
-        provider::ProviderAdmin,
-        store::{AccountStore, AdminStoreError, AdminStoreErrorKind},
-    },
+mod credential_mutation;
+mod error;
+mod publication;
+
+use credential_mutation::{
+    commit_authorization, commit_credential_refresh, commit_credential_rotation,
+    delete_credentials, import_proxy_binding, pending_authorization,
+    publish_credentials_and_observe_quota, required_credential, required_plugin_credential,
+    validate_authorization_commit, validate_prepared_import, validate_prepared_rotation,
+    validate_prepared_rotation_facts,
 };
-use gateway_core::{
-    account::ProviderAccountId,
-    routing::{ConfigRevision, ProviderKind},
-    runtime::SnapshotControl,
-};
-
-fn map_store_error(error: AdminStoreError, resource: &'static str) -> AdminError {
-    let kind = match error.kind() {
-        AdminStoreErrorKind::Invalid => AdminErrorKind::Invalid,
-        AdminStoreErrorKind::NotFound => AdminErrorKind::NotFound,
-        AdminStoreErrorKind::StaleRevision
-        | AdminStoreErrorKind::DuplicateName
-        | AdminStoreErrorKind::Conflict => AdminErrorKind::Conflict,
-        AdminStoreErrorKind::Unavailable => AdminErrorKind::Unavailable,
-    };
-    let message = match kind {
-        AdminErrorKind::Invalid => "请求参数不合法",
-        AdminErrorKind::NotFound => "请求的资源不存在",
-        AdminErrorKind::Conflict => "当前资源状态冲突，请刷新后重试",
-        AdminErrorKind::Unavailable => "依赖服务暂不可用",
-        _ => "服务内部错误",
-    };
-    AdminError::new(kind, message).with_source(OperationFailure {
-        operation: resource,
-        source: error.into(),
-    })
-}
-
-fn map_provider_error(
-    error: crate::ports::provider::ProviderAdminError,
-    resource: &'static str,
-) -> AdminError {
-    use crate::ports::provider::ProviderAdminErrorKind;
-
-    let kind = match error.kind() {
-        ProviderAdminErrorKind::Invalid => AdminErrorKind::Invalid,
-        ProviderAdminErrorKind::Unsupported => AdminErrorKind::Invalid,
-        ProviderAdminErrorKind::NotFound => AdminErrorKind::NotFound,
-        ProviderAdminErrorKind::Conflict => AdminErrorKind::Conflict,
-        ProviderAdminErrorKind::Ambiguous => AdminErrorKind::UpstreamResultUnknown,
-        ProviderAdminErrorKind::Unavailable => AdminErrorKind::Unavailable,
-        ProviderAdminErrorKind::CredentialRefreshRequired => AdminErrorKind::Unavailable,
-        ProviderAdminErrorKind::BadGateway => AdminErrorKind::BadGateway,
-        ProviderAdminErrorKind::Internal => AdminErrorKind::Internal,
-    };
-    let message = match kind {
-        AdminErrorKind::Invalid => "Provider 请求不合法",
-        AdminErrorKind::NotFound => "Provider 资源不存在",
-        AdminErrorKind::Conflict => "Provider 资源状态冲突，请刷新后重试",
-        AdminErrorKind::BadGateway => "上游服务请求失败",
-        AdminErrorKind::UpstreamResultUnknown => "上游执行结果未知，请刷新状态后再决定是否重试",
-        AdminErrorKind::Unavailable => "Provider 服务暂不可用",
-        AdminErrorKind::Internal => "服务内部错误",
-        _ => "Provider 操作失败",
-    };
-    AdminError::new(kind, error.public_message().unwrap_or(message)).with_source(OperationFailure {
-        operation: resource,
-        source: error.into(),
-    })
-}
-
-async fn publish_committed(
-    snapshot: &dyn SnapshotControl,
-    revision: crate::model::Revision,
-) -> Result<(), AdminError> {
-    let revision = ConfigRevision::new(revision.get())
-        .map_err(|_| AdminError::internal("已提交的配置版本不合法"))?;
-    snapshot.publish_committed(revision).await;
-    Ok(())
-}
-
-async fn publish_credentials_and_observe_quota(
-    provider: &std::sync::Arc<dyn crate::ports::provider::ProviderAdmin>,
-    snapshot: &dyn SnapshotControl,
-    revision: crate::model::Revision,
-    account_ids: &[ProviderAccountId],
-    request_id: &str,
-) -> Result<(), AdminError> {
-    provider.account_facts_changed(account_ids).await;
-    publish_committed(snapshot, revision).await?;
-
-    // 凭据已经提交；额度只是可重建的观察，不能拖住管理请求或回滚提交结果
-    let provider = provider.clone();
-    let account_ids = account_ids.to_vec();
-    let request_id = request_id.to_owned();
-    tokio::spawn(async move {
-        for account_id in account_ids {
-            if let Err(error) = provider
-                .quota(ProviderQuotaRequest {
-                    account_id: account_id.clone(),
-                    refresh: true,
-                    rolling_usage: None,
-                })
-                .await
-            {
-                tracing::warn!(
-                    request_id,
-                    provider = %provider.provider_kind().as_str(),
-                    account_id = %account_id.as_str(),
-                    quota_error = ?error.kind(),
-                    "initial quota observation failed"
-                );
-            }
-        }
-    });
-    Ok(())
-}
-
-async fn required_credential(
-    accounts: &dyn AccountStore,
-    provider_kind: &ProviderKind,
-    account_id: &ProviderAccountId,
-    resource: &'static str,
-) -> Result<CredentialDetails, AdminError> {
-    accounts
-        .credential_details(provider_kind, account_id)
-        .await
-        .map_err(|error| map_store_error(error, resource))?
-        .ok_or_else(|| AdminError::not_found("Provider 凭据不存在"))
-}
-
-async fn required_plugin_credential(
-    accounts: &dyn AccountStore,
-    account_id: &ProviderAccountId,
-    resource: &'static str,
-) -> Result<CredentialDetails, AdminError> {
-    accounts
-        .credential_details_by_id(account_id)
-        .await
-        .map_err(|error| map_store_error(error, resource))?
-        .ok_or_else(|| AdminError::not_found("Provider 凭据不存在"))
-}
-
-async fn import_proxy_binding(
-    proxies: &dyn crate::ports::proxy::ProxyStore,
-    id: Option<&str>,
-) -> Result<Option<crate::ports::proxy::ProxyImportReservation>, AdminError> {
-    let Some(id) = id else { return Ok(None) };
-    let reservation = proxies
-        .reserve_import(id)
-        .await
-        .map_err(|error| map_store_error(error, "proxy"))?;
-    Ok(Some(reservation))
-}
-
-async fn pending_authorization(
-    accounts: &dyn AccountStore,
-    proxies: &dyn crate::ports::proxy::ProxyStore,
-    provider_kind: &ProviderKind,
-    command: &StartAuthorization,
-    resource: &'static str,
-) -> Result<PendingAuthorizationMutation, AdminError> {
-    let proxy;
-    let mut proxy_id = None;
-    let target = match &command.reauthorization {
-        Some(account_id) => {
-            proxy = required_credential(accounts, provider_kind, account_id, resource)
-                .await?
-                .credential
-                .outbound_proxy;
-            AuthorizationMutationTarget::Reauthorize {
-                account_id: account_id.clone(),
-            }
-        }
-        None => {
-            use crate::model::proxies::AccountProxySelection;
-            proxy = match &command.outbound_proxy {
-                Some(AccountProxySelection::Url(value)) => Some(value.clone()),
-                Some(AccountProxySelection::Saved(id)) => {
-                    let record = proxies
-                        .get(id)
-                        .await
-                        .map_err(|error| map_store_error(error, "proxy"))?;
-                    proxy_id = Some(record.id);
-                    Some(record.proxy)
-                }
-                Some(AccountProxySelection::Direct) | None => None,
-            };
-            AuthorizationMutationTarget::Create {
-                name: command.name.clone(),
-            }
-        }
-    };
-    Ok(PendingAuthorizationMutation::new(
-        provider_kind.clone(),
-        target,
-        crate::model::provider_credentials::AuthorizationOwnerBinding::from_context(
-            &command.context,
-        ),
-    )
-    .with_outbound_proxy(proxy)
-    .with_outbound_proxy_id(proxy_id))
-}
-
-fn validate_prepared_import(
-    provider_kind: &ProviderKind,
-    prepared: &PreparedCredentialImport,
-    _resource: &'static str,
-) -> Result<(), AdminError> {
-    if prepared.provider_kind != *provider_kind
-        || prepared
-            .credentials
-            .iter()
-            .any(|credential| credential.provider_kind != *provider_kind)
-    {
-        return Err(AdminError::conflict("Provider 准备结果与请求范围不一致"));
-    }
-    Ok(())
-}
-
-fn validate_prepared_rotation(
-    account: &crate::model::accounts::AccountRecord,
-    prepared: &PreparedCredentialRotation,
-    _resource: &'static str,
-) -> Result<(), AdminError> {
-    validate_prepared_rotation_facts(account, prepared.facts())
-}
-
-fn validate_prepared_rotation_facts(
-    account: &crate::model::accounts::AccountRecord,
-    facts: &crate::model::provider_credentials::PreparedCredentialRotationFacts,
-) -> Result<(), AdminError> {
-    if facts.account_id.as_str() != account.id.as_str()
-        || facts.provider_kind != account.provider_kind
-    {
-        return Err(AdminError::conflict("Provider 准备结果与当前凭据不一致"));
-    }
-    Ok(())
-}
-
-async fn validate_authorization_commit(
-    provider_kind: &ProviderKind,
-    context: &MutationContext,
-    prepared: PreparedAuthorizationCommit,
-    resource: &'static str,
-) -> Result<PreparedAuthorizationCommit, AdminError> {
-    if prepared.pending.provider_kind() != provider_kind
-        || !prepared.pending.owner_binding().matches_context(context)
-    {
-        let validation_error = AdminError::conflict("OAuth 授权绑定已失效，请重新发起授权");
-        if let Err(error) = prepared.abort().await {
-            tracing::warn!(
-                resource,
-                settlement_error = %error,
-                "OAuth authorization claim release failed after preparation validation"
-            );
-            return Err(error);
-        }
-        return Err(validation_error);
-    }
-    let matches_target = match (prepared.pending.target(), &prepared.credential) {
-        (
-            AuthorizationMutationTarget::Create { .. },
-            PreparedAuthorizationCredential::Create(credential),
-        ) => credential.provider_kind == *provider_kind,
-        (
-            AuthorizationMutationTarget::Reauthorize { account_id },
-            PreparedAuthorizationCredential::Reauthorize(credential),
-        ) => {
-            let facts = credential.facts();
-            facts.provider_kind == *provider_kind && facts.account_id == *account_id
-        }
-        _ => false,
-    };
-    if !matches_target {
-        let validation_error = AdminError::conflict("OAuth 凭据与待提交目标不一致");
-        if let Err(error) = prepared.abort().await {
-            tracing::warn!(
-                resource,
-                settlement_error = %error,
-                "OAuth authorization claim release failed after preparation validation"
-            );
-            return Err(error);
-        }
-        return Err(validation_error);
-    }
-    Ok(prepared)
-}
-
-async fn commit_authorization(
-    accounts: &dyn AccountStore,
-    prepared: PreparedAuthorizationCommit,
-    key: crate::model::provider_credentials::AuthorizationReceiptKey,
-    settings: Option<crate::model::accounts::AccountImportSettings>,
-    context: &MutationContext,
-    resource: &'static str,
-) -> Result<crate::model::provider_credentials::AuthorizationCommitResult, AdminError> {
-    if settings.is_some()
-        && matches!(
-            &prepared.credential,
-            PreparedAuthorizationCredential::Reauthorize(_)
-        )
-    {
-        prepared.abort().await?;
-        return Err(AdminError::invalid("重新授权不能修改账号设置"));
-    }
-    let crate::model::provider_credentials::AuthorizationCommitSettlement {
-        command,
-        credential_guard,
-        authorization_guard,
-    } = prepared.into_commit(settings, key);
-    match accounts.commit_authorization(command, context).await {
-        Ok(result) => {
-            if let Some(guard) = credential_guard
-                && result.newly_committed
-            {
-                guard.finish();
-            }
-            if let Some(guard) = authorization_guard
-                && let Err(error) = guard.commit().await
-            {
-                // 账号与回执已经原子提交，Redis 清理失败不能把确定的成功改写为失败
-                tracing::warn!(resource, settlement_error = %error, "authorization committed but pending cleanup failed");
-            }
-            Ok(result)
-        }
-        Err(error) => {
-            drop(credential_guard);
-            if let Some(guard) = authorization_guard
-                && let Err(settlement_error) = guard.abort().await
-            {
-                tracing::warn!(
-                    resource,
-                    store_error = %error,
-                    settlement_error = %settlement_error,
-                    "OAuth authorization claim release failed after Store commit failure"
-                );
-                return Err(settlement_error);
-            }
-            Err(map_store_error(error, resource))
-        }
-    }
-}
-
-async fn commit_credential_rotation(
-    accounts: &dyn AccountStore,
-    prepared: PreparedCredentialRotation,
-    settings: Option<crate::model::accounts::UpdateAccount>,
-    context: &MutationContext,
-    resource: &'static str,
-) -> Result<CredentialMutationResult, AdminError> {
-    let (facts, guard) = prepared.into_parts();
-    match accounts
-        .commit_credential_rotation(
-            CredentialRotationCommit {
-                prepared: facts,
-                settings,
-            },
-            context,
-        )
-        .await
-    {
-        Ok(result) => {
-            guard.finish();
-            Ok(result)
-        }
-        Err(error) => {
-            drop(guard);
-            Err(map_store_error(error, resource))
-        }
-    }
-}
-
-async fn commit_credential_refresh(
-    accounts: &dyn AccountStore,
-    prepared: PreparedCredentialRotation,
-    context: &MutationContext,
-    resource: &'static str,
-) -> Result<CredentialMutationResult, AdminError> {
-    let (facts, guard) = prepared.into_parts();
-    match accounts
-        .commit_credential_refresh(
-            CredentialRotationCommit {
-                prepared: facts,
-                settings: None,
-            },
-            context,
-        )
-        .await
-    {
-        Ok(result) => {
-            guard.finish();
-            Ok(result)
-        }
-        Err(error) => {
-            drop(guard);
-            Err(map_store_error(error, resource))
-        }
-    }
-}
-
-async fn delete_credentials(
-    accounts: &dyn AccountStore,
-    provider: &dyn ProviderAdmin,
-    command: CredentialDeletion,
-    resource: &'static str,
-) -> Result<CredentialDeletionResult, AdminError> {
-    for account_id in &command.account_ids {
-        required_credential(accounts, provider.provider_kind(), account_id, resource).await?;
-    }
-    let account_ids = command.account_ids;
-    let revision = accounts
-        .delete_accounts(
-            DeleteAccounts {
-                account_ids: account_ids
-                    .iter()
-                    .map(|account_id| account_id.as_str().to_owned())
-                    .collect(),
-            },
-            &command.context,
-        )
-        .await
-        .map_err(|error| map_store_error(error, resource))?;
-    for account_id in &account_ids {
-        provider.account_unavailable(account_id).await;
-    }
-    provider.account_facts_changed(&account_ids).await;
-    Ok(CredentialDeletionResult {
-        config_revision: revision,
-        account_ids,
-    })
-}
-
-// 用例的操作名补充端口资源上下文，底层分类与原始原因仍由来源持有
-#[derive(Debug, thiserror::Error)]
-#[error("admin operation failed: {operation}")]
-struct OperationFailure {
-    operation: &'static str,
-    source: gateway_core::error::ErrorSource,
-}
+use error::{map_provider_error, map_store_error};
+use publication::publish_committed;

backend/crates/gateway-admin/src/use_case/observability.rs

@@ -17,14 +17,14 @@ use crate::{
         AdminError,
         observability::{
             CostCoverage, CurrencyCost, DashboardAccountUsage, DashboardCapacity,
-            DashboardPeriodMetrics, DashboardResult, DecimalAmount, DiagnosticDimension,
-            DiagnosticsItem, DiagnosticsResult, HealthStatus, HealthTimeline, HealthTimelinePoint,
-            OpsErrorPage, OpsErrorQuery, ProviderBillingInput, RequestMetricPoint, RequestMetrics,
-            TimeRange, Trend, TrendKind, TrendPoint, TrendSummary, UsageBilling,
-            UsageCalculatedBillingFact, UsageDetail, UsageFilter, UsageInsights, UsageInsightsCost,
-            UsageInsightsCostPoint, UsageInsightsHealth, UsageInsightsHealthPoint,
-            UsageInsightsPerformance, UsageInsightsPerformancePoint, UsageOverview, UsagePage,
-            UsageQuery, UsageSummary,
+            DashboardPeriodMetrics, DashboardQuery, DashboardResult, DecimalAmount,
+            DiagnosticDimension, DiagnosticsItem, DiagnosticsResult, Granularity, HealthStatus,
+            HealthTimeline, HealthTimelinePoint, OpsErrorPage, OpsErrorQuery, ProviderBillingInput,
+            RequestMetricPoint, RequestMetrics, TimeRange, Trend, TrendKind, TrendPoint,
+            TrendSummary, UsageBilling, UsageCalculatedBillingFact, UsageDetail, UsageFilter,
+            UsageInsights, UsageInsightsCost, UsageInsightsCostPoint, UsageInsightsHealth,
+            UsageInsightsHealthPoint, UsageInsightsPerformance, UsageInsightsPerformancePoint,
+            UsageOverview, UsagePage, UsageQuery, UsageSummary,
         },
         provider_credentials::ProviderQuotaRequest,
     },
@@ -157,12 +157,13 @@ impl DefaultObservabilityService {
         range: TimeRange,
     ) -> Result<DashboardResult, AdminError> {
         let observed_at = Utc::now();
-        let (mut observation, settings, runtime_slots) = futures::try_join!(
-            self.store.dashboard_summary(range, observed_at),
+        let (mut observation, settings) = futures::try_join!(
+            self.store
+                .dashboard_summary(DashboardQuery::new(range), observed_at),
             self.settings.load_runtime_settings(),
-            self.store.dashboard_runtime_slots(observed_at),
         )
         .map_err(|error| map_store_error(error, "dashboard"))?;
+        let runtime_slots = observation.runtime_slots;
         self.enrich_list_records(&mut observation.recent_requests);
         self.enrich_dashboard_quotas(&mut observation.account_usage)
             .await;
@@ -265,7 +266,7 @@ impl ObservabilityService for DefaultObservabilityService {
     ) -> Result<Trend, AdminError> {
         let points = self
             .store
-            .dashboard_trend(range)
+            .dashboard_trend(range, Granularity::for_range(range))
             .await
             .map_err(|error| map_store_error(error, "dashboard trend"))?;
         trend(kind, points)
@@ -319,10 +320,11 @@ impl ObservabilityService for DefaultObservabilityService {
         range: TimeRange,
         filter: UsageFilter,
     ) -> Result<UsageInsights, AdminError> {
+        let granularity = Granularity::for_range(range);
         let aggregates = async {
             futures::try_join!(
                 self.store.usage_summary(range, filter.clone()),
-                self.store.usage_trend(range, filter.clone()),
+                self.store.usage_trend(range, filter.clone(), granularity),
             )
             .map_err(|error| map_store_error(error, "usage insights"))
         };
@@ -331,7 +333,7 @@ impl ObservabilityService for DefaultObservabilityService {
             recover_standard_costs(
                 &self.providers,
                 self.store
-                    .usage_calculated_billing_facts(range, filter.clone()),
+                    .usage_calculated_billing_facts(range, filter.clone(), granularity),
             ),
         )?;
         build_usage_insights(overview, trend, standard_costs)
@@ -345,7 +347,7 @@ impl ObservabilityService for DefaultObservabilityService {
     ) -> Result<DiagnosticsResult, AdminError> {
         let observation = self
             .store
-            .usage_diagnostics(range, filter, dimension)
+            .usage_diagnostics(range, filter, dimension, 100)
             .await
             .map_err(|error| map_store_error(error, "usage diagnostics"))?;
         let items = observation

backend/crates/gateway-admin/src/use_case/plugins/instances.rs

@@ -2,7 +2,10 @@
 
 use std::collections::{BTreeMap, BTreeSet};
 
-use gateway_core::{policy::ClientApiKeyId, routing::AccountGroupId};
+use gateway_core::{
+    policy::ClientApiKeyId,
+    routing::{AccountGroupId, extensions::ExtensionSetReference},
+};
 use secrecy::ExposeSecret;
 
 use super::PluginsService;
@@ -41,14 +44,11 @@ impl PluginsService {
             .map_err(|error| map_store_error(error, "plugin"))?;
         let warning = match self
             .inspector
-            .inspect(artifact.archive.clone(), Some(digest.to_owned()))
+            .compatibility_warning(artifact.archive, digest.to_owned())
             .await
         {
-            Ok(_) => PackageStatus {
-                warning: self
-                    .inspector
-                    .compatibility_warning(artifact.archive, digest.to_owned())
-                    .await?,
+            Ok(warning) => PackageStatus {
+                warning,
                 load_error: None,
             },
             Err(error) if error.kind() == AdminErrorKind::Invalid => PackageStatus {
@@ -368,25 +368,8 @@ impl PluginsService {
             )
             .await
             .map_err(|e| map_store_error(e, "plugin"))?;
-        let prepared = if result.config_revision == candidate_revision
-            && result.instance.revision == candidate_revision
-        {
-            prepared
-        } else {
-            snapshot.config_revision = result.config_revision;
-            snapshot
-                .instances
-                .retain(|item| item.id != result.instance.id);
-            snapshot.instances.push(result.instance.clone());
-            apply_replacements(&mut snapshot, &replacements);
-            self.preparation.prepare(snapshot).await?
-        };
-        self.preparation
-            .activate_state(&prepared, &result.instance)
+        self.publish_instance(snapshot, prepared, &result, &replacements)
             .await?;
-        // prepared 的强引用跨过提交与唯一发布入口，防止候选在被读取之前回收
-        publish_committed(self.snapshots.as_ref(), result.config_revision).await?;
-        drop(prepared);
         Ok(result)
     }
 
@@ -463,48 +446,45 @@ impl PluginsService {
                     .await);
             }
         };
-        if let Err(error) = self
-            .preparation
-            .migrate_state(&migration_prepared, transition.clone())
-            .await
-        {
-            return Err(self
-                .recover_failed_state_transition(
-                    Some(&transition.id),
-                    &previous,
-                    &previous_state,
-                    disabled_revision,
+        // 迁移开始后的所有提交前失败都走同一恢复出口；提交成功后不再回滚旧配置
+        let committed = async {
+            self.preparation
+                .migrate_state(&migration_prepared, transition.clone())
+                .await?;
+            let mut current = self
+                .store
+                .load_instances()
+                .await
+                .map_err(|error| map_store_error(error, "plugin"))?;
+            if current.config_revision != expected_revision {
+                return Err(AdminError::conflict("迁移期间插件配置已发生变化"));
+            }
+            let candidate_revision = next_revision(expected_revision)?;
+            target.revision = candidate_revision;
+            current.instances.retain(|item| item.id != target.id);
+            current.instances.push(target.clone());
+            current.config_revision = candidate_revision;
+            apply_replacements(&mut current, replacements);
+            let prepared = self.preparation.prepare(current.clone()).await?;
+            let result = self
+                .store
+                .save_instance_replacing(
+                    target,
+                    expected_revision,
+                    PluginStateCommit {
+                        configuration: target_state,
+                        transition_id: Some(transition.id.clone()),
+                    },
+                    replacements,
                     context,
-                    error,
                 )
-                .await);
+                .await
+                .map_err(|error| map_store_error(error, "plugin"))?;
+            Ok((current, prepared, result))
         }
-
-        let mut current = self
-            .store
-            .load_instances()
-            .await
-            .map_err(|error| map_store_error(error, "plugin"))?;
-        if current.config_revision != expected_revision {
-            return Err(self
-                .recover_failed_state_transition(
-                    Some(&transition.id),
-                    &previous,
-                    &previous_state,
-                    disabled_revision,
-                    context,
-                    AdminError::conflict("迁移期间插件配置已发生变化"),
-                )
-                .await);
-        }
-        let candidate_revision = next_revision(expected_revision)?;
-        target.revision = candidate_revision;
-        current.instances.retain(|item| item.id != target.id);
-        current.instances.push(target.clone());
-        current.config_revision = candidate_revision;
-        apply_replacements(&mut current, replacements);
-        let prepared = match self.preparation.prepare(current.clone()).await {
-            Ok(prepared) => prepared,
+        .await;
+        let (current, prepared, result) = match committed {
+            Ok(committed) => committed,
             Err(error) => {
                 return Err(self
                     .recover_failed_state_transition(
@@ -518,53 +498,9 @@ impl PluginsService {
                     .await);
             }
         };
-        let result = match self
-            .store
-            .save_instance_replacing(
-                target,
-                expected_revision,
-                PluginStateCommit {
-                    configuration: target_state,
-                    transition_id: Some(transition.id.clone()),
-                },
-                replacements,
-                context,
-            )
-            .await
-        {
-            Ok(result) => result,
-            Err(error) => {
-                return Err(self
-                    .recover_failed_state_transition(
-                        Some(&transition.id),
-                        &previous,
-                        &previous_state,
-                        disabled_revision,
-                        context,
-                        map_store_error(error, "plugin"),
-                    )
-                    .await);
-            }
-        };
-        let prepared = if result.config_revision == candidate_revision
-            && result.instance.revision == candidate_revision
-        {
-            prepared
-        } else {
-            current.config_revision = result.config_revision;
-            current
-                .instances
-                .retain(|item| item.id != result.instance.id);
-            current.instances.push(result.instance.clone());
-            apply_replacements(&mut current, replacements);
-            self.preparation.prepare(current).await?
-        };
-        self.preparation
-            .activate_state(&prepared, &result.instance)
+        self.publish_instance(current, prepared, &result, replacements)
             .await?;
-        publish_committed(self.snapshots.as_ref(), result.config_revision).await?;
         drop(migration_prepared);
-        drop(prepared);
         Ok(result)
     }
 
@@ -615,22 +551,8 @@ impl PluginsService {
                 )
                 .await
                 .map_err(|error| map_store_error(error, "plugin"))?;
-            let prepared = if mutation.config_revision == candidate_revision
-                && mutation.instance.revision == candidate_revision
-            {
-                prepared
-            } else {
-                snapshot.config_revision = mutation.config_revision;
-                snapshot
-                    .instances
-                    .retain(|item| item.id != mutation.instance.id);
-                snapshot.instances.push(mutation.instance.clone());
-                self.preparation.prepare(snapshot).await?
-            };
-            self.preparation
-                .activate_state(&prepared, &mutation.instance)
+            self.publish_instance(snapshot, prepared, &mutation, &[])
                 .await?;
-            publish_committed(self.snapshots.as_ref(), mutation.config_revision).await?;
             Ok::<(), AdminError>(())
         }
         .await;
@@ -646,6 +568,35 @@ impl PluginsService {
         }
     }
 
+    /// 以事务返回的版本激活并发布，候选强引用必须覆盖整个提交后阶段
+    async fn publish_instance(
+        &self,
+        mut snapshot: PluginInstanceSnapshot,
+        prepared: ExtensionSetReference,
+        mutation: &PluginInstanceMutation,
+        replacements: &[PluginInstanceReplacement],
+    ) -> Result<(), AdminError> {
+        let prepared = if mutation.config_revision == snapshot.config_revision
+            && mutation.instance.revision == snapshot.config_revision
+        {
+            prepared
+        } else {
+            snapshot.config_revision = mutation.config_revision;
+            snapshot
+                .instances
+                .retain(|item| item.id != mutation.instance.id);
+            snapshot.instances.push(mutation.instance.clone());
+            apply_replacements(&mut snapshot, replacements);
+            self.preparation.prepare(snapshot).await?
+        };
+        self.preparation
+            .activate_state(&prepared, &mutation.instance)
+            .await?;
+        publish_committed(self.snapshots.as_ref(), mutation.config_revision).await?;
+        drop(prepared);
+        Ok(())
+    }
+
     pub async fn delete_instance(
         &self,
         id: &str,

backend/crates/gateway-admin/src/use_case/plugins/management.rs

@@ -2,7 +2,8 @@
 
 use std::sync::Arc;
 
-use gateway_core::runtime::{RuntimeSnapshotHandle, extensions::ExtensionSetReference};
+use gateway_core::routing::extensions::ExtensionSetReference;
+use gateway_core::runtime::RuntimeSnapshotHandle;
 
 use crate::{
     model::{

backend/crates/gateway-admin/src/use_case/publication.rs

@@ -0,0 +1,14 @@
+//! 管理事务提交后的运行快照发布入口
+
+use crate::model::AdminError;
+use gateway_core::{routing::ConfigRevision, runtime::SnapshotControl};
+
+pub(super) async fn publish_committed(
+    snapshot: &dyn SnapshotControl,
+    revision: crate::model::Revision,
+) -> Result<(), AdminError> {
+    let revision = ConfigRevision::new(revision.get())
+        .map_err(|_| AdminError::internal("已提交的配置版本不合法"))?;
+    snapshot.publish_committed(revision).await;
+    Ok(())
+}

backend/crates/gateway-admin/tests/backup/support.rs

@@ -519,9 +519,16 @@ impl DatabaseDumpPort for FakeDumpPort {
 
     async fn cleanup_staging(
         &self,
-        _backup_id: &str,
+        backup_id: &str,
     ) -> Result<(), gateway_admin::model::backup::BackupError> {
-        Ok(())
+        match std::fs::remove_file(self.dir.path().join(format!("{backup_id}.dump"))) {
+            Ok(()) => Ok(()),
+            Err(error) if error.kind() == std::io::ErrorKind::NotFound => Ok(()),
+            Err(_) => Err(gateway_admin::model::backup::BackupError::new(
+                gateway_admin::model::backup::code::PG_DUMP_FAILED,
+                "cleanup failed".to_owned(),
+            )),
+        }
     }
 }
 

backend/crates/gateway-admin/tests/backup/task.rs

@@ -55,6 +55,7 @@ async fn backup_task_runs_full_dump_upload_verify_pipeline() {
         .object(&seed.object_key)
         .expect("uploaded object exists");
     assert_eq!(stored.as_slice(), DUMP_CONTENT);
+    assert!(dump.inspect_staging(&seed.id).await.unwrap().is_none());
 }
 
 #[tokio::test]
@@ -259,3 +260,67 @@ async fn timezone_change_rebases_future_cursor_without_running_old_due_backup()
     task.run_cycle(&CancellationToken::new()).await.unwrap();
     assert!(repository.all_records().is_empty());
 }
+
+#[tokio::test]
+async fn completed_recovery_removes_staging_for_dumping_and_uploading_tasks() {
+    for (status, remote_exists) in [
+        (BackupStatus::Dumping, false),
+        (BackupStatus::Uploading, false),
+        (BackupStatus::Uploading, true),
+    ] {
+        let repository = Arc::new(FakeBackupRepository::new(configured_settings()));
+        let dump = Arc::new(FakeDumpPort::new());
+        let object_store = Arc::new(FakeObjectStore::new());
+        let seed = BackupRecordSeed {
+            id: backup_id("recovery-cleanup"),
+            trigger_kind: BackupTriggerKind::Manual,
+            scheduled_at: None,
+            object_key: "recovery.dump".to_owned(),
+            expires_at: None,
+        };
+        repository.insert_backup_record(seed.clone()).await.unwrap();
+        let artifact = dump
+            .dump(DumpRequest {
+                backup_id: seed.id.clone(),
+                cancellation: CancellationToken::new(),
+            })
+            .await
+            .unwrap();
+        repository.force_status(&seed.id, BackupStatus::Dumping);
+        if status == BackupStatus::Uploading {
+            repository
+                .transition_status(
+                    &seed.id,
+                    gateway_admin::model::backup::BackupStatusTransition::try_new(
+                        BackupStatus::Dumping,
+                        status,
+                    )
+                    .unwrap(),
+                    gateway_admin::ports::backup::StatusTransitionUpdate {
+                        size_bytes: Some(artifact.size_bytes),
+                        sha256: Some(artifact.sha256),
+                        ..Default::default()
+                    },
+                    Utc::now(),
+                )
+                .await
+                .unwrap();
+        }
+        if remote_exists {
+            object_store
+                .objects
+                .lock()
+                .unwrap()
+                .insert(seed.object_key.clone(), DUMP_CONTENT.to_vec());
+        }
+        BackupTask::new(repository.clone(), dump.clone(), object_store)
+            .run_cycle(&CancellationToken::new())
+            .await
+            .unwrap();
+        assert_eq!(repository.all_records()[0].status, BackupStatus::Completed);
+        assert!(
+            dump.inspect_staging(&seed.id).await.unwrap().is_none(),
+            "completed recovery must remove staging: {status:?}"
+        );
+    }
+}

backend/crates/gateway-admin/tests/main.rs

@@ -70,12 +70,6 @@ fn request_outcome_filter_should_preserve_known_and_bounded_other_values() {
     assert!(RequestOutcome::new("future\nstate").is_err());
 }
 
-#[test]
-fn time_range_should_accept_exactly_366_days() {
-    let end = Utc::now();
-    assert!(TimeRange::new(end - TimeDelta::days(366), end).is_ok());
-}
-
 #[test]
 fn time_range_should_reject_more_than_366_days() {
     let end = Utc::now();

backend/crates/gateway-admin/tests/use_case/account_groups.rs

@@ -10,12 +10,13 @@ use async_trait::async_trait;
 use chrono::{DateTime, Duration, Utc};
 use gateway_admin::{
     model::{
-        MutationContext, PageSize, Revision,
+        AdminErrorKind, MutationContext, PageSize, Revision,
         account_groups::{
             AccountGroupAccountSummary, AccountGroupCapacity, AccountGroupColor,
-            AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation, AccountGroupPage,
-            AccountGroupRecord, AccountGroupUsage, DeleteAccountGroup, NewAccountGroup,
-            SetAccountGroupEnabled, UpdateAccountGroup,
+            AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation,
+            AccountGroupOptionsPage, AccountGroupPage, AccountGroupRecord, AccountGroupRef,
+            AccountGroupUsage, DeleteAccountGroup, NewAccountGroup, SetAccountGroupEnabled,
+            UpdateAccountGroup,
         },
         accounts::AccountRuntimeSnapshot,
         observability::DecimalAmount,
@@ -90,6 +91,62 @@ async fn group_query_service_enriches_only_current_page_members_with_runtime_fac
     );
 }
 
+#[tokio::test]
+async fn group_options_query_should_skip_member_and_runtime_enrichment() {
+    let groups = Arc::new(FakeGroupStore::default());
+    let runtime = Arc::new(FakeRuntimeStore::default());
+    let service = AdminHarness::new()
+        .account_groups(groups.clone())
+        .account_runtime(runtime.clone())
+        .build()
+        .await;
+
+    let page = service
+        .account_groups()
+        .list_options(AccountGroupListQuery {
+            page: 1,
+            page_size: PageSize::new(20).expect("page size"),
+            search: None,
+            enabled: None,
+        })
+        .await
+        .expect("list account group options");
+
+    assert_eq!(page.items[0].id.as_str(), GROUP_ID);
+    assert!(
+        groups
+            .requested_groups
+            .lock()
+            .expect("requested groups")
+            .is_empty()
+    );
+    assert!(
+        runtime
+            .requested_accounts
+            .lock()
+            .expect("requested accounts")
+            .is_empty()
+    );
+}
+
+#[tokio::test]
+async fn group_options_query_should_preserve_store_failures() {
+    let service = AdminHarness::new().build().await;
+
+    let error = service
+        .account_groups()
+        .list_options(AccountGroupListQuery {
+            page: 1,
+            page_size: PageSize::new(20).expect("page size"),
+            search: None,
+            enabled: None,
+        })
+        .await
+        .expect_err("unavailable options store must fail");
+
+    assert_eq!(error.kind(), AdminErrorKind::Unavailable);
+}
+
 #[derive(Default)]
 struct FakeGroupStore {
     requested_groups: Mutex<Vec<String>>,
@@ -144,6 +201,24 @@ impl AccountGroupStore for FakeGroupStore {
         })
     }
 
+    async fn list_account_group_options(
+        &self,
+        query: AccountGroupListQuery,
+    ) -> AdminStoreResult<AccountGroupOptionsPage> {
+        Ok(AccountGroupOptionsPage {
+            config_revision: Revision::new(1).expect("revision"),
+            items: vec![AccountGroupRef {
+                id: group_id(),
+                name: "Primary".to_owned(),
+                color: AccountGroupColor::parse("#2563EBFF").expect("color"),
+                enabled: true,
+            }],
+            total: 1,
+            page: query.page,
+            page_size: query.page_size.get(),
+        })
+    }
+
     async fn load_account_group_members(
         &self,
         group_ids: &[AccountGroupId],
@@ -300,7 +375,7 @@ fn unused() -> AdminStoreError {
 
 #[tokio::test]
 async fn invalid_group_fields_are_rejected_before_store_mutations() {
-    use gateway_admin::model::{AdminErrorKind, MutationActor, account_groups::CreateAccountGroup};
+    use gateway_admin::model::{MutationActor, account_groups::CreateAccountGroup};
     let service = AdminHarness::new()
         .account_groups(Arc::new(FakeGroupStore::default()))
         .build()

backend/crates/gateway-admin/tests/use_case/accounts.rs

@@ -1538,10 +1538,13 @@ async fn plugin_account_save_rejects_stale_authentication_kind_before_commit() {
     let events = events();
     let provider = FakeProviderAdmin::new("openai", events.clone());
     let store = FakeAccountStore::new("openai", events.clone());
-    let services = accounts_service(provider, store.clone()).await;
+    let access = gateway_admin::initialize_plugin_accounts(
+        ProviderAdminRegistry::new([provider as Arc<dyn ProviderAdmin>]).unwrap(),
+        store.clone(),
+        Arc::new(RecordingPluginAccountPublication(events.clone())),
+    );
 
-    let error = services
-        .plugin_accounts_handle()
+    let error = access
         .save(
             PreparedPluginAccountSave::Replace {
                 facts: plugin_rotation_facts(&account_record("openai")),

backend/crates/gateway-admin/tests/use_case/freeze_recovery.rs

@@ -5,7 +5,6 @@ use std::sync::{Arc, Mutex};
 
 use async_trait::async_trait;
 use chrono::{DateTime, Duration as TimeDelta, Utc};
-use gateway_admin::freeze_recovery::{FreezeRecoveryDeps, FreezeRecoveryTask};
 use gateway_admin::model::MutationContext;
 use gateway_admin::model::accounts::{AccountFreeze, AccountRuntimeSnapshot};
 use gateway_admin::model::settings::{
@@ -18,7 +17,9 @@ use gateway_core::engine::probe::{
     AccountProbe, AccountProbeError, AccountProbeRequest, AccountProbeResult,
 };
 use gateway_core::lifecycle::CancellationToken;
-use gateway_core::task::{ScheduledTask as _, WorkerCycleContext, WorkerId, WorkerKind};
+use gateway_core::task::{
+    ScheduledTask, WorkerContribution, WorkerCycleContext, WorkerId, WorkerKind, WorkerRunnable,
+};
 
 use super::AdminHarness;
 use super::accounts::{FakeAccountStore, FakeProviderAdmin, account_record, events, revision};
@@ -246,7 +247,7 @@ async fn recovery_task(
     settings: RuntimeSettings,
     runtime: Arc<FreezeRuntimeStore>,
     probe: Arc<dyn AccountProbe>,
-) -> (FreezeRecoveryTask, Arc<FakeAccountStore>) {
+) -> (Box<dyn ScheduledTask>, Arc<FakeAccountStore>) {
     recovery_task_with_store(settings, runtime, probe).await
 }
 
@@ -254,9 +255,9 @@ async fn recovery_task_with_store(
     settings: RuntimeSettings,
     runtime: Arc<FreezeRuntimeStore>,
     probe: Arc<dyn AccountProbe>,
-) -> (FreezeRecoveryTask, Arc<FakeAccountStore>) {
+) -> (Box<dyn ScheduledTask>, Arc<FakeAccountStore>) {
     let store = FakeAccountStore::new("openai", events());
-    let services = AdminHarness::new()
+    let bundle = AdminHarness::new()
         .accounts(Arc::clone(&store) as Arc<dyn AccountStore>)
         .account_runtime(Arc::clone(&runtime) as Arc<dyn AccountRuntimeStore>)
         .settings(Arc::new(FreezeSettingsStore {
@@ -264,20 +265,32 @@ async fn recovery_task_with_store(
         }))
         .provider(FakeProviderAdmin::new("openai", events()))
         .probe(probe)
-        .build()
+        .build_bundle()
         .await;
-    let task = FreezeRecoveryTask::new(FreezeRecoveryDeps {
-        accounts: services.accounts_handle(),
-        store: store.clone(),
-        runtime,
-        settings: Arc::new(FreezeSettingsStore {
-            settings: settings.clone(),
-        }),
-    });
+    let task = freeze_worker(bundle);
     (task, store)
 }
 
-async fn run_cycle(task: &FreezeRecoveryTask) {
+fn freeze_worker(mut bundle: gateway_admin::AdminBundle) -> Box<dyn ScheduledTask> {
+    bundle
+        .take_worker_contributions()
+        .into_iter()
+        .find_map(|contribution| {
+            let WorkerContribution::Registration(registration) = contribution else {
+                return None;
+            };
+            if registration.id.kind() != WorkerKind::AccountFreezeRecovery {
+                return None;
+            }
+            match registration.runnable {
+                WorkerRunnable::Scheduled { task, .. } => Some(task),
+                WorkerRunnable::Daemon { .. } => None,
+            }
+        })
+        .expect("registered freeze recovery worker")
+}
+
+async fn run_cycle(task: &dyn ScheduledTask) {
     let worker = WorkerId::try_new(WorkerKind::AccountFreezeRecovery, "test").expect("worker id");
     let context = WorkerCycleContext::new(worker, None, CancellationToken::new());
     task.run_cycle(context)
@@ -311,7 +324,7 @@ async fn probe_success_clears_freeze() {
     )
     .await;
 
-    run_cycle(&task).await;
+    run_cycle(task.as_ref()).await;
 
     assert_eq!(runtime.cleared(), vec!["acct_test".to_owned()]);
     assert!(runtime.extended().is_empty());
@@ -327,7 +340,7 @@ async fn probe_failure_postpones_freeze() {
     )
     .await;
 
-    run_cycle(&task).await;
+    run_cycle(task.as_ref()).await;
 
     assert!(runtime.cleared().is_empty());
     let extended = runtime.extended();
@@ -351,7 +364,7 @@ async fn disabled_policy_releases_due_freeze_without_probing() {
     )
     .await;
 
-    run_cycle(&task).await;
+    run_cycle(task.as_ref()).await;
 
     assert_eq!(runtime.cleared(), vec!["acct_test".to_owned()]);
     assert!(runtime.extended().is_empty());
@@ -370,7 +383,7 @@ async fn adaptive_concurrency_lowers_limit_to_observed_peak() {
     )
     .await;
 
-    run_cycle(&task).await;
+    run_cycle(task.as_ref()).await;
 
     // 峰值 4 × 0.8 = 3（不低于下限 2），低于全局默认 5，应下调到 3
     assert_eq!(
@@ -395,7 +408,7 @@ async fn adaptive_concurrency_never_raises_limit() {
     )
     .await;
 
-    run_cycle(&task).await;
+    run_cycle(task.as_ref()).await;
 
     // 峰值 40 的目标 32 高于全局默认 5；只降不升，不下发任何更新
     assert!(store.update_commands().is_empty());
@@ -415,7 +428,7 @@ async fn probe_skips_freezes_far_from_expiry() {
     )
     .await;
 
-    run_cycle(&task).await;
+    run_cycle(task.as_ref()).await;
 
     assert!(runtime.cleared().is_empty());
     assert!(runtime.extended().is_empty());
@@ -431,7 +444,7 @@ async fn disabled_accounts_are_skipped() {
     let mut record = account_record("openai");
     record.enabled = false;
     let store = FakeAccountStore::with_account(record, events());
-    let services = AdminHarness::new()
+    let bundle = AdminHarness::new()
         .accounts(Arc::clone(&store) as Arc<dyn AccountStore>)
         .account_runtime(Arc::clone(&runtime) as Arc<dyn AccountRuntimeStore>)
         .settings(Arc::new(FreezeSettingsStore {
@@ -439,18 +452,11 @@ async fn disabled_accounts_are_skipped() {
         }))
         .provider(FakeProviderAdmin::new("openai", events()))
         .probe(Arc::new(SuccessfulProbe))
-        .build()
+        .build_bundle()
         .await;
-    let task = FreezeRecoveryTask::new(FreezeRecoveryDeps {
-        accounts: services.accounts_handle(),
-        store: store.clone(),
-        runtime: Arc::clone(&runtime) as Arc<dyn AccountRuntimeStore>,
-        settings: Arc::new(FreezeSettingsStore {
-            settings: runtime_settings(true, true, true),
-        }),
-    });
+    let task = freeze_worker(bundle);
 
-    run_cycle(&task).await;
+    run_cycle(task.as_ref()).await;
 
     assert!(runtime.cleared().is_empty());
     assert!(runtime.extended().is_empty());

backend/crates/gateway-admin/tests/use_case/mod.rs

@@ -29,8 +29,9 @@ use gateway_admin::{
     model::{
         MutationContext, Revision,
         account_groups::{
-            AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation, AccountGroupPage,
-            DeleteAccountGroup, NewAccountGroup, SetAccountGroupEnabled, UpdateAccountGroup,
+            AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation,
+            AccountGroupOptionsPage, AccountGroupPage, DeleteAccountGroup, NewAccountGroup,
+            SetAccountGroupEnabled, UpdateAccountGroup,
         },
         accounts::{
             AccountListQuery, AccountPage, AccountRuntimeSnapshot, AccountUpdateResult,
@@ -44,10 +45,10 @@ use gateway_admin::{
             NewClientKey, SetClientKeyEnabled, UpdateClientKey,
         },
         observability::{
-            DashboardDesktopRelease, DashboardObservation, DashboardWireAttribute,
+            DashboardDesktopRelease, DashboardObservation, DashboardQuery, DashboardWireAttribute,
             DashboardWireProfile, DashboardWireTarget, DesktopReleaseStatus, DiagnosticDimension,
-            DiagnosticsObservation, OpsErrorPage, OpsErrorQuery, RequestMetricPoint, TimeRange,
-            UsageDetail, UsageFilter, UsageOverview, UsagePage, UsageQuery,
+            DiagnosticsObservation, Granularity, OpsErrorPage, OpsErrorQuery, RequestMetricPoint,
+            TimeRange, UsageDetail, UsageFilter, UsageOverview, UsagePage, UsageQuery,
         },
         provider_credentials::{
             AuthorizationCommit, AuthorizationStarted, CompleteAuthorization, CredentialDetails,
@@ -107,7 +108,7 @@ pub(super) struct AdminHarness {
     plugin_store: Arc<dyn gateway_admin::ports::plugins::PluginStore>,
     plugin_inspector: Arc<dyn gateway_admin::ports::plugins::PluginPackageInspector>,
     client_key_verifier: Arc<dyn ClientKeyVerifier>,
-    service_middleware: gateway_admin::service::PlanSource,
+    service_middleware: gateway_admin::public_service::PlanSource,
 }
 
 impl AdminHarness {
@@ -194,7 +195,10 @@ impl AdminHarness {
         self
     }
 
-    pub(super) fn service_middleware(mut self, source: gateway_admin::service::PlanSource) -> Self {
+    pub(super) fn service_middleware(
+        mut self,
+        source: gateway_admin::public_service::PlanSource,
+    ) -> Self {
         self.service_middleware = source;
         self
     }
@@ -421,6 +425,13 @@ impl AccountGroupStore for UnavailableAccountGroupStore {
         Err(unavailable("account groups"))
     }
 
+    async fn list_account_group_options(
+        &self,
+        _: AccountGroupListQuery,
+    ) -> AdminStoreResult<AccountGroupOptionsPage> {
+        Err(unavailable("account group options"))
+    }
+
     async fn load_account_group_members(
         &self,
         _: &[gateway_core::routing::AccountGroupId],
@@ -738,13 +749,17 @@ impl ClientKeyStore for UnavailableStore {
 impl ObservabilityStore for UnavailableStore {
     async fn dashboard_summary(
         &self,
-        _: TimeRange,
+        _: DashboardQuery,
         _: DateTime<Utc>,
     ) -> AdminStoreResult<DashboardObservation> {
         Err(unavailable("dashboard"))
     }
 
-    async fn dashboard_trend(&self, _: TimeRange) -> AdminStoreResult<Vec<RequestMetricPoint>> {
+    async fn dashboard_trend(
+        &self,
+        _: TimeRange,
+        _: Granularity,
+    ) -> AdminStoreResult<Vec<RequestMetricPoint>> {
         Err(unavailable("dashboard trend"))
     }
 
@@ -752,6 +767,7 @@ impl ObservabilityStore for UnavailableStore {
         &self,
         _: TimeRange,
         _: UsageFilter,
+        _: Granularity,
     ) -> AdminStoreResult<Vec<RequestMetricPoint>> {
         Err(unavailable("usage trend"))
     }
@@ -760,6 +776,7 @@ impl ObservabilityStore for UnavailableStore {
         &self,
         _: TimeRange,
         _: UsageFilter,
+        _: Granularity,
     ) -> gateway_admin::ports::store::UsageCalculatedBillingStream<'_> {
         Box::pin(futures::stream::once(async {
             Err(unavailable("usage billing facts"))
@@ -783,6 +800,7 @@ impl ObservabilityStore for UnavailableStore {
         _: TimeRange,
         _: UsageFilter,
         _: DiagnosticDimension,
+        _: u16,
     ) -> AdminStoreResult<DiagnosticsObservation> {
         Err(unavailable("usage diagnostics"))
     }

backend/crates/gateway-admin/tests/use_case/observability.rs

@@ -18,7 +18,7 @@ use gateway_admin::{
         MutationContext, Revision,
         observability::{
             AccountPoolMetrics, AttemptMetrics, CostCoverage, CurrencyCost, DashboardAccountUsage,
-            DashboardObservation, DashboardRuntimeSlots, DiagnosticDimension,
+            DashboardObservation, DashboardQuery, DashboardRuntimeSlots, DiagnosticDimension,
             DiagnosticObservation, DiagnosticsObservation, Granularity, HealthStatus,
             LatencyPercentiles, ObservabilityPageSize, OpsErrorPage, OpsErrorQuery,
             PercentileMilliseconds, RequestMetricPoint, RequestMetrics, TimeRange, TrendKind,
@@ -282,7 +282,7 @@ async fn dashboard_capacity_distinguishes_unlimited_inheritance_finite_overrides
 }
 
 #[tokio::test]
-async fn dashboard_summary_should_share_one_current_observation_time_across_runtime_facts() {
+async fn dashboard_summary_should_observe_current_accounts_for_a_historical_range() {
     let historical_end = Utc::now() - Duration::hours(1);
     let store = Arc::new(FixtureObservabilityStore::new(observation_range(
         historical_end,
@@ -295,8 +295,7 @@ async fn dashboard_summary_should_share_one_current_observation_time_across_runt
         .await
         .expect("dashboard summary");
 
-    let (summary_observed_at, slots_observed_at) = store.observed_times();
-    assert_eq!(summary_observed_at, slots_observed_at);
+    let summary_observed_at = *store.summary_observed_at.lock().unwrap();
     assert!(summary_observed_at.is_some_and(|value| value > historical_end));
 }
 
@@ -789,7 +788,8 @@ struct FixtureObservabilityStore {
     diagnostics: Mutex<DiagnosticsObservation>,
     runtime_slots: Mutex<Option<DashboardRuntimeSlots>>,
     summary_observed_at: Mutex<Option<DateTime<Utc>>>,
-    slots_observed_at: Mutex<Option<DateTime<Utc>>>,
+    summary_query: Mutex<Option<DashboardQuery>>,
+    usage_granularities: Mutex<Vec<Granularity>>,
     usage_records: Mutex<Vec<UsageListRecord>>,
     account_usage: Mutex<Vec<DashboardAccountUsage>>,
     dashboard_delay: Mutex<StdDuration>,
@@ -811,7 +811,8 @@ impl FixtureObservabilityStore {
             diagnostics: Mutex::new(DiagnosticsObservation::default()),
             runtime_slots: Mutex::new(None),
             summary_observed_at: Mutex::new(None),
-            slots_observed_at: Mutex::new(None),
+            summary_query: Mutex::new(None),
+            usage_granularities: Mutex::new(Vec::new()),
             usage_records: Mutex::new(Vec::new()),
             account_usage: Mutex::new(Vec::new()),
             dashboard_delay: Mutex::new(StdDuration::ZERO),
@@ -850,16 +851,6 @@ impl FixtureObservabilityStore {
         *self.runtime_slots.lock().expect("runtime slots") = runtime_slots;
     }
 
-    fn observed_times(&self) -> (Option<DateTime<Utc>>, Option<DateTime<Utc>>) {
-        (
-            *self
-                .summary_observed_at
-                .lock()
-                .expect("summary observed at"),
-            *self.slots_observed_at.lock().expect("slots observed at"),
-        )
-    }
-
     fn replace_usage_records(&self, records: Vec<UsageListRecord>) {
         *self.usage_records.lock().expect("usage records") = records;
     }
@@ -869,10 +860,11 @@ impl FixtureObservabilityStore {
 impl ObservabilityStore for FixtureObservabilityStore {
     async fn dashboard_summary(
         &self,
-        range: TimeRange,
+        query: DashboardQuery,
         observed_at: DateTime<Utc>,
     ) -> AdminStoreResult<DashboardObservation> {
         self.dashboard_summary_calls.fetch_add(1, Ordering::Relaxed);
+        *self.summary_query.lock().unwrap() = Some(query.clone());
         *self
             .summary_observed_at
             .lock()
@@ -882,24 +874,21 @@ impl ObservabilityStore for FixtureObservabilityStore {
             tokio::time::sleep(dashboard_delay).await;
         }
         Ok(DashboardObservation {
-            range,
+            range: query.range,
             totals: Default::default(),
             provider_accounts: AccountPoolMetrics::default(),
+            runtime_slots: *self.runtime_slots.lock().expect("runtime slots"),
             trend: self.trend.lock().expect("trend").clone(),
             account_usage: self.account_usage.lock().expect("account usage").clone(),
             recent_requests: Vec::new(),
         })
     }
 
-    async fn dashboard_runtime_slots(
+    async fn dashboard_trend(
         &self,
-        observed_at: DateTime<Utc>,
-    ) -> AdminStoreResult<Option<DashboardRuntimeSlots>> {
-        *self.slots_observed_at.lock().expect("slots observed at") = Some(observed_at);
-        Ok(*self.runtime_slots.lock().expect("runtime slots"))
-    }
-
-    async fn dashboard_trend(&self, _: TimeRange) -> AdminStoreResult<Vec<RequestMetricPoint>> {
+        _: TimeRange,
+        _: Granularity,
+    ) -> AdminStoreResult<Vec<RequestMetricPoint>> {
         Ok(self.trend.lock().expect("trend").clone())
     }
 
@@ -907,7 +896,9 @@ impl ObservabilityStore for FixtureObservabilityStore {
         &self,
         _: TimeRange,
         _: UsageFilter,
+        granularity: Granularity,
     ) -> AdminStoreResult<Vec<RequestMetricPoint>> {
+        self.usage_granularities.lock().unwrap().push(granularity);
         Ok(self.trend.lock().expect("trend").clone())
     }
 
@@ -915,7 +906,9 @@ impl ObservabilityStore for FixtureObservabilityStore {
         &self,
         _: TimeRange,
         _: UsageFilter,
+        granularity: Granularity,
     ) -> gateway_admin::ports::store::UsageCalculatedBillingStream<'_> {
+        self.usage_granularities.lock().unwrap().push(granularity);
         let facts = self
             .calculated_billing_facts
             .lock()
@@ -954,6 +947,7 @@ impl ObservabilityStore for FixtureObservabilityStore {
         _: TimeRange,
         _: UsageFilter,
         _: DiagnosticDimension,
+        _: u16,
     ) -> AdminStoreResult<DiagnosticsObservation> {
         Ok(self.diagnostics.lock().expect("diagnostics").clone())
     }
@@ -1144,6 +1138,13 @@ fn total_record(
         first_text_ms: None,
         first_token_ms: None,
         provider_processing_ms: None,
+        upstream_response_ms: None,
+        upstream_api_overhead_ms: None,
+        upstream_engine_ms: None,
+        upstream_engine_iapi_ttft_ms: None,
+        upstream_engine_service_ttft_ms: None,
+        upstream_engine_iapi_tbt_ms: None,
+        upstream_engine_service_tbt_ms: None,
         latency_ms: Some(100),
         admission_decision_ms: None,
         account_selection_wait_ms: None,
@@ -1195,3 +1196,41 @@ fn quarter_hour_start(value: DateTime<Utc>) -> DateTime<Utc> {
     let elapsed = value.timestamp().rem_euclid(15 * 60);
     value - Duration::seconds(elapsed) - Duration::nanoseconds(i64::from(value.nanosecond()))
 }
+
+#[tokio::test]
+async fn admin_selects_dashboard_defaults_and_matching_usage_buckets_at_range_boundaries() {
+    let now = Utc::now();
+    for (duration, expected) in [
+        (Duration::days(2), Granularity::FifteenMinutes),
+        (Duration::days(2) + Duration::seconds(1), Granularity::Hour),
+        (Duration::days(31), Granularity::Hour),
+        (Duration::days(31) + Duration::seconds(1), Granularity::Day),
+    ] {
+        let range = TimeRange::new(now - duration, now).unwrap();
+        let store = Arc::new(FixtureObservabilityStore::new(range));
+        let services = observability_services(store.clone()).await;
+        services
+            .observability()
+            .dashboard_summary(range, TrendKind::Usage)
+            .await
+            .unwrap();
+        let query = store.summary_query.lock().unwrap().clone().unwrap();
+        assert_eq!(query.range, range);
+        assert_eq!(query.granularity, expected);
+        assert_eq!(query.account_limit, 4);
+        assert_eq!(query.recent_request_limit, 10);
+        assert_eq!(
+            query.recent_request_filter.outcome,
+            Some(gateway_admin::model::observability::RequestOutcome::Succeeded)
+        );
+        services
+            .observability()
+            .usage_insights(range, UsageFilter::default())
+            .await
+            .unwrap();
+        assert_eq!(
+            *store.usage_granularities.lock().unwrap(),
+            vec![expected, expected]
+        );
+    }
+}

backend/crates/gateway-admin/tests/use_case/plugins/artifacts.rs

@@ -24,12 +24,10 @@ use gateway_admin::{
         store::{AdminStoreError, AdminStoreErrorKind, AdminStoreResult},
     },
 };
+use gateway_core::routing::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference};
 use gateway_core::{
     routing::ConfigRevision,
-    runtime::{
-        RuntimeSnapshotHandle, SnapshotControl,
-        extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference},
-    },
+    runtime::{RuntimeSnapshotHandle, SnapshotControl},
 };
 use serde_json::json;
 
@@ -117,7 +115,7 @@ impl gateway_admin::ports::plugins::PluginRuntimeDiagnostics for Fixture {
         &self,
         _: &gateway_admin::model::plugins::instances::PluginInstanceSnapshot,
         _: Option<u64>,
-        _: Option<&gateway_core::runtime::extensions::ExtensionSetReference>,
+        _: Option<&gateway_core::routing::extensions::ExtensionSetReference>,
     ) -> Option<
         std::collections::BTreeMap<
             String,
@@ -132,7 +130,7 @@ impl gateway_admin::ports::plugins::PluginRuntimeDiagnostics for Fixture {
 impl gateway_admin::ports::plugins::PluginStateLifecycle for Fixture {
     async fn activate_state(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::instances::PluginInstance,
     ) -> Result<(), AdminError> {
         Ok(())
@@ -142,7 +140,7 @@ impl gateway_admin::ports::plugins::PluginStateLifecycle for Fixture {
     }
     async fn migrate_state(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: gateway_admin::model::plugins::state::PluginStateTransition,
     ) -> Result<(), AdminError> {
         panic!("unexpected state migration")

backend/crates/gateway-admin/tests/use_case/plugins/instances.rs

@@ -40,12 +40,10 @@ use gateway_admin::{
         store::{AdminStoreError, AdminStoreErrorKind, AdminStoreResult},
     },
 };
+use gateway_core::routing::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference};
 use gateway_core::{
     routing::ConfigRevision,
-    runtime::{
-        RuntimeSnapshotHandle, SnapshotControl,
-        extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference},
-    },
+    runtime::{RuntimeSnapshotHandle, SnapshotControl},
 };
 use serde_json::json;
 
@@ -87,9 +85,10 @@ impl gateway_admin::ports::plugins::PluginPackageInspector for LifecycleFixture
     }
     async fn compatibility_warning(
         &self,
-        _: Arc<[u8]>,
-        _: String,
+        archive: Arc<[u8]>,
+        digest: String,
     ) -> Result<Option<String>, AdminError> {
+        self.inspect(archive, Some(digest)).await?;
         Ok(self.data.lock().unwrap().compatibility_warning.clone())
     }
     async fn inspect(
@@ -132,6 +131,9 @@ struct FixtureData {
     saves: Vec<SavedInstance>,
     history: BTreeMap<String, PluginVersionConfiguration>,
     aborted: usize,
+    transition_active: bool,
+    fail_reload_after_migration: bool,
+    fail_next_load: bool,
     activated: usize,
     quiesced: usize,
     diagnostics: Option<BTreeMap<String, PluginInstanceRuntime>>,
@@ -185,6 +187,9 @@ impl LifecycleFixture {
                     },
                 )]),
                 aborted: 0,
+                transition_active: false,
+                fail_reload_after_migration: false,
+                fail_next_load: false,
                 activated: 0,
                 quiesced: 0,
                 diagnostics: None,
@@ -237,6 +242,9 @@ impl LifecycleFixture {
             .instances
             .retain(|current| current.id != instance.id);
         data.snapshot.instances.push(instance.clone());
+        if transition_id.is_some() {
+            data.transition_active = false;
+        }
         data.saves.push(SavedInstance {
             artifact_sha256: instance.artifact_sha256.clone(),
             enabled: instance.enabled,
@@ -336,7 +344,11 @@ impl gateway_admin::ports::plugins::PluginStateLifecycle for LifecycleFixture {
         _: PluginStateTransition,
     ) -> Result<(), AdminError> {
         match self.behavior {
-            MigrationBehavior::Succeed => Ok(()),
+            MigrationBehavior::Succeed => {
+                let mut data = self.data.lock().unwrap();
+                data.fail_next_load = std::mem::take(&mut data.fail_reload_after_migration);
+                Ok(())
+            }
             MigrationBehavior::Fail => Err(AdminError::invalid("migration fixture failed")),
             MigrationBehavior::ConcurrentChangeThenFail => {
                 let mut data = self.data.lock().unwrap();
@@ -400,9 +412,10 @@ impl PluginStateStore for LifecycleFixture {
         artifact_sha256: &str,
         _: PluginStateConfiguration,
     ) -> PluginStateStoreResult<PluginStateTransition> {
-        let data = self.data.lock().unwrap();
+        let mut data = self.data.lock().unwrap();
         let current = &data.snapshot.instances[0];
-        if current.enabled
+        if data.transition_active
+            || current.enabled
             || current.id != instance_id
             || current.revision != expected_instance_revision
             || artifact_sha256 != NEW_ARTIFACT
@@ -411,6 +424,7 @@ impl PluginStateStore for LifecycleFixture {
                 PluginStateStoreErrorKind::Conflict,
             ));
         }
+        data.transition_active = true;
         Ok(PluginStateTransition {
             id: "00000000-0000-7000-8000-000000000002".into(),
             instance_id: instance_id.into(),
@@ -440,7 +454,9 @@ impl PluginStateStore for LifecycleFixture {
     }
 
     async fn abort_transition(&self, _: &str) -> PluginStateStoreResult<()> {
-        self.data.lock().unwrap().aborted += 1;
+        let mut data = self.data.lock().unwrap();
+        data.aborted += 1;
+        data.transition_active = false;
         Ok(())
     }
 }
@@ -465,7 +481,11 @@ impl PluginStore for LifecycleFixture {
             }))
     }
     async fn load_instances(&self) -> AdminStoreResult<PluginInstanceSnapshot> {
-        Ok(self.snapshot())
+        let mut data = self.data.lock().unwrap();
+        if std::mem::take(&mut data.fail_next_load) {
+            return Err(admin_error(AdminStoreErrorKind::Unavailable));
+        }
+        Ok(data.snapshot.clone())
     }
 
     async fn load_version_configuration(
@@ -1048,6 +1068,36 @@ async fn failed_state_migration_restores_the_previous_enabled_version_with_cas()
     assert_eq!(published.revisions.lock().unwrap().as_slice(), [11, 12]);
 }
 
+#[tokio::test]
+async fn reload_failure_after_migration_aborts_restores_and_allows_retry() {
+    let fixture = LifecycleFixture::new(MigrationBehavior::Succeed);
+    fixture.data.lock().unwrap().fail_reload_after_migration = true;
+    let published = Arc::new(Published::default());
+    let service = service(fixture.clone(), published.clone());
+    let id = "00000000-0000-7000-8000-000000000001";
+    let error = service
+        .configure_instance(Some(id), input(), &context())
+        .await
+        .err()
+        .unwrap();
+    assert_eq!(error.kind(), AdminErrorKind::Unavailable);
+    {
+        let data = fixture.data.lock().unwrap();
+        assert_eq!(data.aborted, 1);
+        assert!(!data.transition_active);
+        assert!(data.snapshot.instances[0].enabled);
+        assert_eq!(data.snapshot.instances[0].artifact_sha256, OLD_ARTIFACT);
+        assert_eq!(*published.revisions.lock().unwrap(), [11, 12]);
+    }
+    let result = service
+        .configure_instance(Some(id), input(), &context())
+        .await
+        .unwrap();
+    assert_eq!(result.instance.artifact_sha256, NEW_ARTIFACT);
+    assert!(result.instance.enabled);
+    assert!(!fixture.data.lock().unwrap().transition_active);
+}
+
 #[tokio::test]
 async fn failed_state_migration_never_overwrites_a_concurrent_configuration_change() {
     let fixture = LifecycleFixture::new(MigrationBehavior::ConcurrentChangeThenFail);

backend/crates/gateway-admin/tests/use_case/plugins/mod.rs

@@ -111,7 +111,7 @@ impl gateway_admin::ports::proxy::ProxyStore for TestPluginPorts {
 impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPorts {
     async fn validate_target(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
     ) -> Result<(), AdminError> {
         Err(AdminError::not_found("unused plugin fixture"))
@@ -119,7 +119,7 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
 
     async fn start_callback(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         _: gateway_admin::model::plugins::management::StartPluginManagementCallback,
         _: &gateway_admin::model::auth::AdminRequestContext,
@@ -129,7 +129,7 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
     }
     async fn callback(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         _: &str,
         _: gateway_admin::model::plugins::management::PluginManagementRequest,
@@ -139,14 +139,14 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
     }
     async fn views(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
     ) -> Result<Vec<gateway_admin::model::plugins::management::PluginManagementView>, AdminError>
     {
         Ok(Vec::new())
     }
     async fn resource(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         _: &str,
         _: bool,
@@ -156,7 +156,7 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
     }
     async fn handle(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         _: gateway_admin::model::plugins::management::PluginManagementRequest,
     ) -> Result<gateway_admin::model::plugins::management::PluginManagementResponse, AdminError>
@@ -171,7 +171,7 @@ impl gateway_admin::ports::plugins::PluginRuntimeDiagnostics for TestPluginPorts
         &self,
         _: &gateway_admin::model::plugins::instances::PluginInstanceSnapshot,
         _: Option<u64>,
-        _: Option<&gateway_core::runtime::extensions::ExtensionSetReference>,
+        _: Option<&gateway_core::routing::extensions::ExtensionSetReference>,
     ) -> Option<
         std::collections::BTreeMap<
             String,
@@ -186,7 +186,7 @@ impl gateway_admin::ports::plugins::PluginRuntimeDiagnostics for TestPluginPorts
 impl gateway_admin::ports::plugins::PluginStateLifecycle for TestPluginPorts {
     async fn activate_state(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::instances::PluginInstance,
     ) -> Result<(), AdminError> {
         panic!("unexpected state activation")
@@ -196,7 +196,7 @@ impl gateway_admin::ports::plugins::PluginStateLifecycle for TestPluginPorts {
     }
     async fn migrate_state(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: gateway_admin::model::plugins::state::PluginStateTransition,
     ) -> Result<(), AdminError> {
         panic!("unexpected state migration")
@@ -222,7 +222,7 @@ impl gateway_admin::ports::plugins::PluginPreparation for TestPluginPorts {
     async fn prepare(
         &self,
         _: gateway_admin::model::plugins::instances::PluginInstanceSnapshot,
-    ) -> Result<gateway_core::runtime::extensions::ExtensionSetReference, AdminError> {
+    ) -> Result<gateway_core::routing::extensions::ExtensionSetReference, AdminError> {
         Err(AdminError::invalid("unused plugin fixture"))
     }
 }

backend/crates/gateway-admin/tests/use_case/service_contract.rs

@@ -247,7 +247,7 @@ fn sdk_settings_contract_matches_host_declarations() {
             _ => None,
         })
         .collect();
-    let register = source(root, "src/service/settings.rs")
+    let register = source(root, "src/public_service/settings.rs")
         .items
         .into_iter()
         .find_map(|item| match item {
@@ -293,7 +293,7 @@ fn sdk_settings_contract_matches_host_declarations() {
     let generated = quote! {
         //! 宿主设置服务的操作标识与请求、响应数据合同
         //!
-        //! 从宿主设置类型与 service/settings.rs 生成；更新命令见 SDK 维护说明
+        //! 从宿主设置类型与 public_service/settings.rs 生成；更新命令见 SDK 维护说明
 
         use super::Operation;
         use serde::{Deserialize, Serialize};

backend/crates/gateway-admin/tests/use_case/settings.rs

@@ -28,7 +28,7 @@ struct PricingSettingsStore {
 
 #[tokio::test]
 async fn explicit_service_calls_freeze_empty_plan_and_cancel_before_entering_terminal() {
-    use gateway_admin::service::{Origin, Plan, scope};
+    use gateway_admin::public_service::{Origin, Plan, scope};
     use std::sync::atomic::{AtomicUsize, Ordering};
 
     let resolutions = Arc::new(AtomicUsize::new(0));

backend/crates/gateway-api/src/admin/account_groups.rs

@@ -16,8 +16,9 @@ use gateway_admin::model::{
     PageSize,
     account_groups::{
         AccountGroupAccountSummary, AccountGroupCapacity, AccountGroupColor, AccountGroupListQuery,
-        AccountGroupMutation, AccountGroupPage, AccountGroupRecord, AccountGroupUsage,
-        CreateAccountGroup, DeleteAccountGroup, SetAccountGroupEnabled, UpdateAccountGroup,
+        AccountGroupMutation, AccountGroupOptionsPage, AccountGroupPage, AccountGroupRecord,
+        AccountGroupUsage, CreateAccountGroup, DeleteAccountGroup, SetAccountGroupEnabled,
+        UpdateAccountGroup,
     },
 };
 use gateway_core::{account::FastMode, routing::AccountGroupId};
@@ -212,6 +213,38 @@ struct AccountGroupPageData {
     config_revision: u64,
 }
 
+#[derive(Debug, Clone, Serialize)]
+#[serde(rename_all = "camelCase")]
+struct AccountGroupOptionsPageData {
+    items: Vec<AccountGroupRefView>,
+    page: PageMeta,
+    config_revision: u64,
+}
+
+impl From<AccountGroupOptionsPage> for AccountGroupOptionsPageData {
+    fn from(page: AccountGroupOptionsPage) -> Self {
+        let total_pages = if page.total == 0 {
+            0
+        } else {
+            page.total.div_ceil(u64::from(page.page_size))
+        };
+        Self {
+            items: page
+                .items
+                .into_iter()
+                .map(AccountGroupRefView::from)
+                .collect(),
+            page: PageMeta::new(
+                page.page,
+                u32::from(page.page_size),
+                page.total,
+                u32::try_from(total_pages).unwrap_or(u32::MAX),
+            ),
+            config_revision: page.config_revision.get(),
+        }
+    }
+}
+
 impl From<(AccountGroupPage, crate::time::TimePresenter)> for AccountGroupPageData {
     fn from((page, time): (AccountGroupPage, crate::time::TimePresenter)) -> Self {
         let total_pages = if page.total == 0 {
@@ -263,6 +296,7 @@ where
 {
     Router::new()
         .route("/api/admin/account-groups", get(list::<S>))
+        .route("/api/admin/account-groups/options", get(list_options::<S>))
         .route("/api/admin/account-groups/create", post(create::<S>))
         .route("/api/admin/account-groups/update", post(update::<S>))
         .route("/api/admin/account-groups/enable", post(enable::<S>))
@@ -270,6 +304,26 @@ where
         .route("/api/admin/account-groups/delete", post(delete::<S>))
 }
 
+async fn list_options<S>(
+    _auth: AdminAuth,
+    State(state): State<S>,
+    AdminQuery(query): AdminQuery<ListAccountGroupsQuery>,
+) -> Result<impl IntoResponse, AdminError>
+where
+    S: SessionState + Send + Sync,
+{
+    let result = state
+        .admin_services()
+        .account_groups()
+        .list_options(query.into_command().map_err(map_wire_error)?)
+        .await
+        .map_err(map_service_error)?;
+    Ok(AdminResponse::new(
+        StatusCode::OK,
+        AdminEnvelope::ok(AccountGroupOptionsPageData::from(result)),
+    ))
+}
+
 async fn list<S>(
     _auth: AdminAuth,
     State(state): State<S>,

backend/crates/gateway-api/src/admin/observability/presenter.rs

@@ -307,8 +307,14 @@ pub(crate) fn usage_list_record_view(
             first_event_ms: record.first_event_ms,
             first_reasoning_ms: record.first_reasoning_ms,
             first_text_ms: record.first_text_ms,
-            first_token_ms: record.first_token_ms,
             openai_processing_ms: record.provider_processing_ms,
+            upstream_response_ms: record.upstream_response_ms,
+            upstream_api_overhead_ms: record.upstream_api_overhead_ms,
+            upstream_engine_ms: record.upstream_engine_ms,
+            upstream_engine_iapi_ttft_ms: record.upstream_engine_iapi_ttft_ms,
+            upstream_engine_service_ttft_ms: record.upstream_engine_service_ttft_ms,
+            upstream_engine_iapi_tbt_ms: record.upstream_engine_iapi_tbt_ms,
+            upstream_engine_service_tbt_ms: record.upstream_engine_service_tbt_ms,
         },
         first_token_latency_ms: record.first_token_ms,
         latency_ms: record.latency_ms,
@@ -345,7 +351,6 @@ pub(crate) fn usage_record_view(
         .error_message
         .clone()
         .unwrap_or_else(|| outcome.clone());
-    let first_token_display = display_duration(record.first_token_ms);
     let latency_display = display_duration(record.latency_ms);
     let cost_coverage = match record.cost_source.as_str() {
         "provider_reported" | "calculated" => CostCoverageView {
@@ -409,15 +414,20 @@ pub(crate) fn usage_record_view(
             first_event_ms: record.first_event_ms,
             first_reasoning_ms: record.first_reasoning_ms,
             first_text_ms: record.first_text_ms,
-            first_token_ms: record.first_token_ms,
             openai_processing_ms: record.provider_processing_ms,
+            upstream_response_ms: record.upstream_response_ms,
+            upstream_api_overhead_ms: record.upstream_api_overhead_ms,
+            upstream_engine_ms: record.upstream_engine_ms,
+            upstream_engine_iapi_ttft_ms: record.upstream_engine_iapi_ttft_ms,
+            upstream_engine_service_ttft_ms: record.upstream_engine_service_ttft_ms,
+            upstream_engine_iapi_tbt_ms: record.upstream_engine_iapi_tbt_ms,
+            upstream_engine_service_tbt_ms: record.upstream_engine_service_tbt_ms,
         },
         attempt_index: None,
         attempt_count: u64::from(record.attempt_count),
         response_id: record.client_response_id,
         upstream_request_id: record.upstream_request_id,
         latency_ms: record.latency_ms,
-        first_token_ms: record.first_token_ms,
         input_tokens: record.input_tokens,
         output_tokens: record.output_tokens,
         cached_tokens: record.cached_tokens,
@@ -441,7 +451,6 @@ pub(crate) fn usage_record_view(
         costs,
         cost_coverage,
         first_token_latency_ms: record.first_token_ms,
-        first_token_latency_ms_display: first_token_display,
         latency_ms_display: latency_display,
         logical_outcome: outcome,
     }
@@ -756,7 +765,7 @@ pub(crate) fn trend_summary_view(
                 ratio: None,
             },
             TrendSummaryView {
-                label: "吞吐 P50".to_owned(),
+                label: "速率 P50".to_owned(),
                 value: summary
                     .minimum_output_throughput_p50
                     .map_or_else(|| "—".to_owned(), |value| format!("{value} tok/s")),
@@ -904,6 +913,7 @@ pub(crate) fn dashboard_view(
         rotation_strategy,
     } = result;
     let domain::DashboardObservation {
+        runtime_slots: _,
         range,
         totals,
         provider_accounts,

backend/crates/gateway-api/src/admin/observability/query.rs

@@ -54,15 +54,7 @@ pub struct UsageQuery {
 impl UsageQuery {
     /// 校验 Element Plus 风格的页码分页字段
     pub fn validate_pagination(&self) -> Result<(u32, u16), WireValidationError> {
-        let current_page = self.current_page.unwrap_or(1);
-        if current_page == 0 {
-            return Err(WireValidationError::new("currentPage"));
-        }
-        let page_size = self.page_size.unwrap_or(DEFAULT_PAGE_SIZE);
-        if DomainPageSize::new(page_size).is_err() {
-            return Err(WireValidationError::new("pageSize"));
-        }
-        Ok((current_page, page_size))
+        validate_pagination(self.current_page, self.page_size)
     }
 }
 
@@ -131,15 +123,7 @@ pub struct OpsQuery {
 impl OpsQuery {
     /// 校验 Element Plus 风格的页码分页字段
     pub fn validate_pagination(&self) -> Result<(u32, u16), WireValidationError> {
-        let current_page = self.current_page.unwrap_or(1);
-        if current_page == 0 {
-            return Err(WireValidationError::new("currentPage"));
-        }
-        let page_size = self.page_size.unwrap_or(DEFAULT_PAGE_SIZE);
-        if DomainPageSize::new(page_size).is_err() {
-            return Err(WireValidationError::new("pageSize"));
-        }
-        Ok((current_page, page_size))
+        validate_pagination(self.current_page, self.page_size)
     }
 }
 
@@ -396,3 +380,18 @@ pub(crate) fn non_empty(value: Option<String>) -> Option<String> {
         .map(|value| value.trim().to_owned())
         .filter(|value| !value.is_empty())
 }
+
+fn validate_pagination(
+    current_page: Option<u32>,
+    page_size: Option<u16>,
+) -> Result<(u32, u16), WireValidationError> {
+    let current_page = current_page.unwrap_or(1);
+    if current_page == 0 {
+        return Err(WireValidationError::new("currentPage"));
+    }
+    let page_size = page_size.unwrap_or(DEFAULT_PAGE_SIZE);
+    if DomainPageSize::new(page_size).is_err() {
+        return Err(WireValidationError::new("pageSize"));
+    }
+    Ok((current_page, page_size))
+}

backend/crates/gateway-api/src/admin/observability/wire.rs

@@ -160,7 +160,6 @@ pub struct UsageRecordView {
     pub response_id: Option<String>,
     pub upstream_request_id: Option<String>,
     pub latency_ms: Option<u64>,
-    pub first_token_ms: Option<u64>,
     pub input_tokens: Option<u64>,
     pub output_tokens: Option<u64>,
     pub cached_tokens: Option<u64>,
@@ -185,7 +184,6 @@ pub struct UsageRecordView {
     pub costs: Vec<CostView>,
     pub cost_coverage: CostCoverageView,
     pub first_token_latency_ms: Option<u64>,
-    pub first_token_latency_ms_display: String,
     pub latency_ms_display: String,
     pub logical_outcome: String,
 }
@@ -222,9 +220,21 @@ pub struct UsageLatencyDetailsView {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub first_text_ms: Option<u64>,
     #[serde(skip_serializing_if = "Option::is_none")]
-    pub first_token_ms: Option<u64>,
-    #[serde(skip_serializing_if = "Option::is_none")]
     pub openai_processing_ms: Option<u64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub upstream_response_ms: Option<u64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub upstream_api_overhead_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub upstream_engine_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub upstream_engine_iapi_ttft_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub upstream_engine_service_ttft_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub upstream_engine_iapi_tbt_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    pub upstream_engine_service_tbt_ms: Option<f64>,
 }
 
 /// 单次上游尝试展示

backend/crates/gateway-api/src/key_usage/presenter.rs

@@ -201,6 +201,20 @@ pub(super) struct RecordView {
 #[derive(Default, Serialize)]
 #[serde(rename_all = "camelCase")]
 struct OutputTimingView {
+    #[serde(skip_serializing_if = "Option::is_none")]
+    upstream_response_ms: Option<u64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    upstream_api_overhead_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    upstream_engine_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    upstream_engine_iapi_ttft_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    upstream_engine_service_ttft_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    upstream_engine_iapi_tbt_ms: Option<f64>,
+    #[serde(skip_serializing_if = "Option::is_none")]
+    upstream_engine_service_tbt_ms: Option<f64>,
     #[serde(skip_serializing_if = "Option::is_none")]
     first_event_ms: Option<u64>,
     #[serde(skip_serializing_if = "Option::is_none")]
@@ -226,6 +240,13 @@ fn success_record(value: UsageListRecord, time: crate::time::TimePresenter) -> R
         latency_ms: value.latency_ms,
         first_token_latency_ms: value.first_token_ms,
         latency_details: OutputTimingView {
+            upstream_response_ms: value.upstream_response_ms,
+            upstream_api_overhead_ms: value.upstream_api_overhead_ms,
+            upstream_engine_ms: value.upstream_engine_ms,
+            upstream_engine_iapi_ttft_ms: value.upstream_engine_iapi_ttft_ms,
+            upstream_engine_service_ttft_ms: value.upstream_engine_service_ttft_ms,
+            upstream_engine_iapi_tbt_ms: value.upstream_engine_iapi_tbt_ms,
+            upstream_engine_service_tbt_ms: value.upstream_engine_service_tbt_ms,
             first_event_ms: value.first_event_ms,
             first_reasoning_ms: value.first_reasoning_ms,
             first_text_ms: value.first_text_ms,

backend/crates/gateway-api/src/lib.rs

@@ -138,7 +138,7 @@ impl ApiBundle {
     }
 
     /// 内部调用使用同一总路由；组合根持有强引用，Runtime 只保存 Weak
-    pub fn dispatcher(&self) -> Arc<dyn gateway_core::middleware::http::Dispatcher> {
+    pub fn dispatcher(&self) -> Arc<dyn gateway_core::engine::middleware::http::Dispatcher> {
         Arc::new(middleware::RouterDispatcher {
             router: middleware::wrap(
                 self.router.clone(),

backend/crates/gateway-api/src/middleware/mod.rs

@@ -17,23 +17,21 @@ use axum::{
     http::{HeaderName, HeaderValue, Request, StatusCode},
     response::{IntoResponse as _, Response},
 };
-use gateway_core::{
-    engine::middleware::FrozenMiddlewarePlan,
-    lifecycle::CancellationToken,
-    middleware::{
-        compose,
-        http::{self as contract, Settings},
-    },
-};
+use gateway_core::engine::middleware::FrozenMiddlewarePlan;
+use gateway_core::engine::middleware::http as contract;
+use gateway_core::engine::middleware::http::Settings;
+use gateway_core::lifecycle::CancellationToken;
+use gateway_core::middleware::compose;
 use http_body::{Body as HttpBody, Frame, SizeHint};
 use http_body_util::BodyExt as _;
 use tokio::time::Instant;
 use tower::ServiceExt as _;
 
 tokio::task_local! { static CURRENT: contract::Context; }
-tokio::task_local! { static SETTINGS: Option<gateway_core::settings::RequestSettings>; }
+tokio::task_local! { static SETTINGS: Option<gateway_core::routing::request_settings::RequestSettings>; }
 
-pub(crate) fn current_settings() -> Option<gateway_core::settings::RequestSettings> {
+pub(crate) fn current_settings() -> Option<gateway_core::routing::request_settings::RequestSettings>
+{
     SETTINGS.try_with(Clone::clone).ok().flatten()
 }
 
@@ -52,7 +50,7 @@ pub(crate) async fn scope<T>(
 }
 
 pub(crate) type SettingsSource =
-    Arc<dyn Fn() -> Option<gateway_core::settings::RequestSettings> + Send + Sync>;
+    Arc<dyn Fn() -> Option<gateway_core::routing::request_settings::RequestSettings> + Send + Sync>;
 
 pub(crate) struct RouterDispatcher {
     pub(crate) router: Router,
@@ -60,7 +58,7 @@ pub(crate) struct RouterDispatcher {
 }
 
 impl contract::Dispatcher for RouterDispatcher {
-    fn request_settings(&self) -> Option<gateway_core::settings::RequestSettings> {
+    fn request_settings(&self) -> Option<gateway_core::routing::request_settings::RequestSettings> {
         (self.settings)()
     }
 
@@ -195,12 +193,12 @@ async fn handle(
     });
     request.extensions_mut().insert(guard.clone());
     request.extensions_mut().insert(context.clone());
-    let service_context = gateway_admin::service::Origin {
+    let service_context = gateway_admin::public_service::Origin {
         request_id: context.request_id.clone(),
         call_id: context.call_id.clone(),
         cancellation: context.cancellation.clone(),
         extensions: context.extensions.clone(),
-        plan: gateway_admin::service::Plan::Frozen(plan.clone()),
+        plan: gateway_admin::public_service::Plan::Frozen(plan.clone()),
     };
     request
         .extensions_mut()
@@ -220,7 +218,7 @@ async fn handle(
                     terminal_context,
                     SETTINGS.scope(
                         settings,
-                        gateway_admin::service::scope(
+                        gateway_admin::public_service::scope(
                             service_context,
                             dispatch(router, request.map(Body::new), started),
                         ),

backend/crates/gateway-api/src/middleware/websocket.rs

@@ -14,15 +14,14 @@ use futures::{
     future::BoxFuture,
     stream::{SplitSink, SplitStream},
 };
-use gateway_core::{
-    engine::middleware::{FrozenMiddlewarePlan, MiddlewareError, MiddlewareHeader},
-    lifecycle::CancellationToken,
-    middleware::{
-        compose,
-        http::{self as http, upgrade},
-        websocket as core,
-    },
-};
+use gateway_core::engine::middleware::FrozenMiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareError;
+use gateway_core::engine::middleware::MiddlewareHeader;
+use gateway_core::engine::middleware::http;
+use gateway_core::engine::middleware::http::upgrade;
+use gateway_core::engine::middleware::websocket as core;
+use gateway_core::lifecycle::CancellationToken;
+use gateway_core::middleware::compose;
 use http_body_util::BodyExt as _;
 
 pub(super) struct Upgrade {

backend/crates/gateway-api/src/openai/auth.rs

@@ -84,30 +84,6 @@ impl ClientApiKeyAuthError {
     }
 }
 
-/// 从请求头提取 Bearer Client API key
-///
-/// # Errors
-///
-/// Header 缺失、Bearer 语法错误或 Key 不能作为 HTTP Bearer 值时返回稳定错误
-pub fn bearer_client_api_key(headers: &HeaderMap) -> Result<&str, ClientApiKeyAuthError> {
-    let raw = headers
-        .get(AUTHORIZATION)
-        .ok_or(ClientApiKeyAuthError::MissingAuthorization)?
-        .to_str()
-        .map_err(|_| ClientApiKeyAuthError::MalformedAuthorization)?;
-    let token = raw
-        .strip_prefix("Bearer ")
-        .ok_or(ClientApiKeyAuthError::MalformedAuthorization)?
-        .trim();
-    if token.is_empty() {
-        return Err(ClientApiKeyAuthError::MalformedAuthorization);
-    }
-    if gateway_core::policy::PlaintextClientApiKey::validate(token).is_err() {
-        return Err(ClientApiKeyAuthError::InvalidKeyFormat);
-    }
-    Ok(token)
-}
-
 pub(crate) async fn authenticate_client(
     service: &OpenAiService,
     headers: &HeaderMap,

backend/crates/gateway-api/src/openai/responses/websocket/connection/frame.rs

@@ -0,0 +1,171 @@
+//! 下游连接的帧、写入结果、退出原因与队列预算
+
+use std::{sync::Arc, time::Duration};
+use thiserror::Error;
+use tokio::sync::OwnedSemaphorePermit;
+
+pub(super) const OUTBOUND_COMMAND_BUFFER: usize = 32;
+pub(super) const INBOUND_EVENT_BUFFER: usize = 32;
+pub(super) const DOWNSTREAM_WRITE_TIMEOUT: Duration = Duration::from_secs(5 * 60);
+pub(super) const DOWNSTREAM_CLOSE_TIMEOUT: Duration = Duration::from_secs(1);
+pub(super) const CONNECTION_MAX_AGE: Duration = Duration::from_secs(60 * 60);
+pub(super) const DOWNSTREAM_PING_INTERVAL: Duration = Duration::from_secs(25);
+
+/// WebSocket pump 的写入与生命周期预算
+#[derive(Clone, Copy)]
+pub struct ConnectionConfig {
+    pub(super) write_timeout: Duration,
+    pub(super) max_age: Duration,
+}
+
+impl ConnectionConfig {
+    /// 生产环境固定预算
+    pub const PRODUCTION: Self = Self {
+        write_timeout: DOWNSTREAM_WRITE_TIMEOUT,
+        max_age: CONNECTION_MAX_AGE,
+    };
+}
+
+/// 业务层可观察的客户端输入；Ping/Pong 始终由 pump 消费
+pub enum ConnectionEvent {
+    Text(String),
+    Binary,
+    Expired,
+    Exited(PumpExitReason),
+}
+
+/// 接收队列与活动期间暂存的请求共用容量；移动帧不释放它占用的名额
+pub(in super::super) struct PendingConnectionEvent {
+    pub(in super::super) event: ConnectionEvent,
+    pub(super) _permit: Option<OwnedSemaphorePermit>,
+}
+
+impl PendingConnectionEvent {
+    pub(super) fn exited(reason: PumpExitReason) -> Self {
+        Self {
+            event: ConnectionEvent::Exited(reason),
+            _permit: None,
+        }
+    }
+}
+
+/// 下游写入阶段；名称刻意使用 write，而不是暗示客户端已消费的 delivery
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum FramePhase {
+    Metadata,
+    First,
+    Data,
+    Terminal,
+    FirstAndTerminal,
+    Error,
+    ConnectionLimit,
+    Close,
+}
+
+impl FramePhase {
+    pub(super) const fn as_str(self) -> &'static str {
+        match self {
+            Self::Metadata => "metadata",
+            Self::First => "first",
+            Self::Data => "data",
+            Self::Terminal => "terminal",
+            Self::FirstAndTerminal => "first_and_terminal",
+            Self::Error => "error",
+            Self::ConnectionLimit => "connection_limit",
+            Self::Close => "close",
+        }
+    }
+
+    pub(super) const fn is_milestone(self) -> bool {
+        matches!(
+            self,
+            Self::First
+                | Self::Terminal
+                | Self::FirstAndTerminal
+                | Self::Error
+                | Self::ConnectionLimit
+                | Self::Close
+        )
+    }
+}
+
+/// 一次下游写入的请求归属与协议阶段
+#[derive(Clone)]
+pub struct WriteContext {
+    pub(super) request_id: Option<Arc<str>>,
+    pub(super) phase: FramePhase,
+}
+
+impl WriteContext {
+    /// 创建归属于某个请求的写入上下文
+    #[must_use]
+    pub fn request(request_id: &Arc<str>, phase: FramePhase) -> Self {
+        Self {
+            request_id: Some(Arc::clone(request_id)),
+            phase,
+        }
+    }
+
+    /// 创建连接级写入上下文
+    #[must_use]
+    pub const fn connection(phase: FramePhase) -> Self {
+        Self {
+            request_id: None,
+            phase,
+        }
+    }
+}
+
+/// WebSocket pump 停止的稳定原因
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum PumpExitReason {
+    ClientClose,
+    PeerEof,
+    ReadError,
+    WriteError,
+    WriteTimeout,
+    LifecycleShutdown,
+    ConnectionMaxAge,
+    CoordinatorDropped,
+    InboundOverload,
+    ServerClose,
+    PumpStopped,
+    MiddlewareFailed,
+}
+
+impl PumpExitReason {
+    pub(in super::super) const fn as_str(self) -> &'static str {
+        match self {
+            Self::ClientClose => "client_close",
+            Self::PeerEof => "peer_eof",
+            Self::ReadError => "read_error",
+            Self::WriteError => "write_error",
+            Self::WriteTimeout => "write_timeout",
+            Self::LifecycleShutdown => "lifecycle_shutdown",
+            Self::ConnectionMaxAge => "connection_max_age",
+            Self::CoordinatorDropped => "coordinator_dropped",
+            Self::InboundOverload => "inbound_overload",
+            Self::ServerClose => "server_close",
+            Self::PumpStopped => "pump_stopped",
+            Self::MiddlewareFailed => "middleware_failed",
+        }
+    }
+}
+
+/// 下游 WebSocket 写入失败
+#[derive(Debug, Error)]
+pub enum ConnectionWriteError {
+    #[error("downstream WebSocket pump is closed")]
+    Closed,
+    #[error("downstream WebSocket write timed out after {timeout:?}")]
+    Timeout { timeout: Duration },
+    #[error("downstream WebSocket transport write failed: {message}")]
+    Transport { message: String },
+}
+
+/// 只有实际 transport 写入才记为 Written；插件丢弃消息不伪造写入成功
+#[derive(Clone, Copy, Debug, PartialEq, Eq)]
+pub enum WriteOutcome {
+    Written,
+    Suppressed,
+}

backend/crates/gateway-api/src/openai/responses/websocket/connection/handle.rs

@@ -0,0 +1,362 @@
+//! 协调层持有的连接句柄与唯一 pump 启动入口
+
+use super::{
+    frame::*,
+    pump::run_pump,
+    state::{ConnectionCommand, ConnectionStats},
+};
+use axum::extract::ws::{CloseFrame, Message, WebSocket, close_code};
+use futures::{Sink, Stream};
+use gateway_core::{
+    engine::middleware::{FrozenMiddlewarePlan, MiddlewareHeader},
+    lifecycle::CancellationToken,
+};
+use std::{
+    collections::VecDeque,
+    fmt,
+    sync::{
+        Arc,
+        atomic::{AtomicBool, Ordering},
+    },
+    time::Duration,
+};
+use tokio::{
+    sync::{mpsc, oneshot},
+    task::JoinHandle,
+    time::{Instant, timeout},
+};
+
+/// 协调层持有的单 owner WebSocket 连接句柄
+pub struct ResponsesWebSocketConnection {
+    connection_id: Arc<str>,
+    opened_at: Instant,
+    expired: Arc<AtomicBool>,
+    commands: Option<mpsc::Sender<ConnectionCommand>>,
+    incoming: mpsc::Receiver<PendingConnectionEvent>,
+    deferred: VecDeque<PendingConnectionEvent>,
+    exited: oneshot::Receiver<PumpExitReason>,
+    pump_task: Option<JoinHandle<()>>,
+    stats: Arc<ConnectionStats>,
+    config: ConnectionConfig,
+    exit_reason: Option<PumpExitReason>,
+}
+
+impl ResponsesWebSocketConnection {
+    pub(in super::super) fn new(
+        socket: WebSocket,
+        connection_id: String,
+        cancellation: CancellationToken,
+        middleware: Option<FrozenMiddlewarePlan>,
+        headers: Arc<[MiddlewareHeader]>,
+    ) -> Self {
+        spawn_connection(
+            socket,
+            Arc::<str>::from(connection_id),
+            cancellation,
+            ConnectionConfig::PRODUCTION,
+            middleware,
+            headers,
+        )
+    }
+
+    pub(in super::super) fn id(&self) -> &str {
+        &self.connection_id
+    }
+
+    /// 返回连接是否已经达到生命周期上限
+    #[must_use]
+    pub fn is_expired(&self) -> bool {
+        self.expired.load(Ordering::Acquire)
+    }
+
+    pub(in super::super) fn age(&self) -> Duration {
+        self.opened_at.elapsed()
+    }
+
+    /// 等待下一个需要业务层处理的客户端事件
+    pub async fn next_event(&mut self) -> Option<ConnectionEvent> {
+        if self.exit_reason.is_none() {
+            match self.exited.try_recv() {
+                Ok(reason) => {
+                    self.exit_reason = Some(reason);
+                }
+                Err(tokio::sync::oneshot::error::TryRecvError::Closed) => {
+                    self.exit_reason = Some(PumpExitReason::PumpStopped);
+                }
+                Err(tokio::sync::oneshot::error::TryRecvError::Empty) => {}
+            }
+        }
+        if let Some(reason) = self.exit_reason {
+            return Some(ConnectionEvent::Exited(reason));
+        }
+        if let Some(event) = self.deferred.pop_front() {
+            return Some(event.event);
+        }
+        self.next_active_event().await.map(|event| event.event)
+    }
+
+    /// 活动响应期间只消费新到的帧，已排队的下一轮请求继续保持串行
+    pub(in super::super) async fn next_active_event(&mut self) -> Option<PendingConnectionEvent> {
+        if let Some(reason) = self.exit_reason {
+            return Some(PendingConnectionEvent::exited(reason));
+        }
+        let event = tokio::select! {
+            biased;
+            reason = &mut self.exited => {
+                Some(PendingConnectionEvent::exited(reason.unwrap_or(PumpExitReason::PumpStopped)))
+            }
+            event = self.incoming.recv() => event,
+        };
+        if let Some(PendingConnectionEvent {
+            event: ConnectionEvent::Exited(reason),
+            ..
+        }) = event.as_ref()
+        {
+            self.exit_reason.get_or_insert(*reason);
+        } else if event.is_none() {
+            self.exit_reason.get_or_insert(PumpExitReason::PumpStopped);
+        }
+        event
+    }
+
+    pub(in super::super) fn defer(&mut self, event: PendingConnectionEvent) {
+        self.deferred.push_back(event);
+    }
+
+    /// 等待连接退出，不消费留给后续串行请求的业务帧
+    pub async fn wait_for_exit(&mut self) -> PumpExitReason {
+        if let Some(reason) = self.exit_reason {
+            return reason;
+        }
+        let reason = (&mut self.exited)
+            .await
+            .unwrap_or(PumpExitReason::PumpStopped);
+        self.exit_reason = Some(reason);
+        reason
+    }
+
+    /// 串行写入文本帧，并等待 pump 确认 transport 写入结果
+    ///
+    /// # Errors
+    ///
+    /// pump 已关闭、写入超时或 transport 失败时返回稳定错误
+    pub async fn send_text(
+        &mut self,
+        payload: String,
+        context: WriteContext,
+    ) -> Result<WriteOutcome, ConnectionWriteError> {
+        self.send(Message::Text(payload.into()), context).await
+    }
+
+    pub(in super::super) async fn close_policy(
+        &mut self,
+        reason: &'static str,
+        request_id: Option<&Arc<str>>,
+    ) {
+        let context = request_id.map_or_else(
+            || WriteContext::connection(FramePhase::Close),
+            |request_id| WriteContext::request(request_id, FramePhase::Close),
+        );
+        self.close(
+            CloseFrame {
+                code: close_code::POLICY,
+                reason: reason.into(),
+            },
+            context,
+            PumpExitReason::ServerClose,
+        )
+        .await;
+    }
+
+    pub(in super::super) async fn close_for_connection_limit(&mut self, reason: &'static str) {
+        self.close(
+            CloseFrame {
+                code: close_code::NORMAL,
+                reason: reason.into(),
+            },
+            WriteContext::connection(FramePhase::Close),
+            PumpExitReason::ConnectionMaxAge,
+        )
+        .await;
+    }
+
+    async fn close(
+        &mut self,
+        frame: CloseFrame,
+        context: WriteContext,
+        exit_reason: PumpExitReason,
+    ) {
+        if self
+            .send(Message::Close(Some(frame)), context)
+            .await
+            .is_ok()
+        {
+            self.exit_reason.get_or_insert(exit_reason);
+        }
+    }
+
+    async fn send(
+        &mut self,
+        message: Message,
+        context: WriteContext,
+    ) -> Result<WriteOutcome, ConnectionWriteError> {
+        let Some(commands) = self.commands.as_ref().cloned() else {
+            return Err(ConnectionWriteError::Closed);
+        };
+        self.stats.observe_command_queue(&commands);
+        let (acknowledged, acknowledgement) = oneshot::channel();
+        let started_at = Instant::now();
+        let request_id = context.request_id.clone();
+        let phase = context.phase;
+        let command = ConnectionCommand {
+            message,
+            context,
+            acknowledged,
+        };
+        let write = async move {
+            commands
+                .send(command)
+                .await
+                .map_err(|_| ConnectionWriteError::Closed)?;
+            acknowledgement
+                .await
+                .map_err(|_| ConnectionWriteError::Closed)?
+        };
+        let result = match timeout(self.config.write_timeout, write).await {
+            Ok(result) => result,
+            Err(_) => Err(ConnectionWriteError::Timeout {
+                timeout: self.config.write_timeout,
+            }),
+        };
+        let duration = started_at.elapsed();
+        match &result {
+            Ok(WriteOutcome::Written) if phase.is_milestone() => tracing::info!(
+                websocket_connection_id = %self.connection_id,
+                request_id = request_id.as_deref().unwrap_or(""),
+                frame_phase = phase.as_str(),
+                write_duration_ms = duration.as_millis(),
+                "Responses WebSocket frame write succeeded"
+            ),
+            Ok(outcome) => tracing::debug!(
+                websocket_connection_id = %self.connection_id,
+                request_id = request_id.as_deref().unwrap_or(""),
+                frame_phase = phase.as_str(),
+                write_duration_ms = duration.as_millis(),
+                ?outcome,
+                "Responses WebSocket frame processed"
+            ),
+            Err(error) => tracing::info!(
+                websocket_connection_id = %self.connection_id,
+                request_id = request_id.as_deref().unwrap_or(""),
+                frame_phase = phase.as_str(),
+                write_duration_ms = duration.as_millis(),
+                error = %error,
+                "Responses WebSocket frame write failed"
+            ),
+        }
+        if let Err(error) = &result {
+            let reason = match error {
+                ConnectionWriteError::Timeout { .. } => PumpExitReason::WriteTimeout,
+                ConnectionWriteError::Transport { .. } => PumpExitReason::WriteError,
+                ConnectionWriteError::Closed => PumpExitReason::PumpStopped,
+            };
+            self.terminate(reason);
+        }
+        result
+    }
+
+    fn terminate(&mut self, reason: PumpExitReason) {
+        self.exit_reason.get_or_insert(reason);
+        self.commands.take();
+        if let Some(task) = self.pump_task.take() {
+            task.abort();
+        }
+    }
+
+    pub(in super::super) fn log_summary(&self, request_count: u64) {
+        let connection_age_ms = self.age().as_millis();
+        let read_idle_ms = connection_age_ms
+            .saturating_sub(u128::from(self.stats.last_read_ms.load(Ordering::Relaxed)));
+        let write_idle_ms = connection_age_ms
+            .saturating_sub(u128::from(self.stats.last_write_ms.load(Ordering::Relaxed)));
+        tracing::info!(
+            websocket_connection_id = %self.connection_id,
+            request_count,
+            connection_age_ms,
+            pump_exit_reason = self
+                .exit_reason
+                .map_or("unknown", PumpExitReason::as_str),
+            command_queue_high_water = self
+                .stats
+                .command_queue_high_water
+                .load(Ordering::Relaxed),
+            command_backpressure_count = self
+                .stats
+                .command_backpressure_count
+                .load(Ordering::Relaxed),
+            ping_received_count = self.stats.ping_received_count.load(Ordering::Relaxed),
+            ping_written_count = self.stats.ping_written_count.load(Ordering::Relaxed),
+            pong_received_count = self.stats.pong_received_count.load(Ordering::Relaxed),
+            read_idle_ms,
+            write_idle_ms,
+            "Responses WebSocket disconnected"
+        );
+    }
+}
+
+impl Drop for ResponsesWebSocketConnection {
+    fn drop(&mut self) {
+        self.commands.take();
+        if let Some(task) = self.pump_task.take() {
+            task.abort();
+        }
+    }
+}
+
+/// 为 WebSocket transport 启动包含冻结中间件计划的单 owner pump
+pub fn spawn_connection<S, E>(
+    socket: S,
+    connection_id: Arc<str>,
+    cancellation: CancellationToken,
+    config: ConnectionConfig,
+    middleware: Option<FrozenMiddlewarePlan>,
+    headers: Arc<[MiddlewareHeader]>,
+) -> ResponsesWebSocketConnection
+where
+    S: Stream<Item = Result<Message, E>> + Sink<Message, Error = E> + Unpin + Send + 'static,
+    E: fmt::Display + Send + 'static,
+{
+    let opened_at = Instant::now();
+    let expired = Arc::new(AtomicBool::new(false));
+    let stats = Arc::new(ConnectionStats::default());
+    let (command_tx, command_rx) = mpsc::channel(OUTBOUND_COMMAND_BUFFER);
+    let (incoming_tx, incoming_rx) = mpsc::channel(INBOUND_EVENT_BUFFER);
+    let (exit_tx, exit_rx) = oneshot::channel();
+    let pump_task = tokio::spawn(run_pump(
+        socket,
+        command_rx,
+        incoming_tx,
+        exit_tx,
+        Arc::clone(&connection_id),
+        cancellation,
+        opened_at,
+        Arc::clone(&expired),
+        Arc::clone(&stats),
+        config,
+        middleware,
+        headers,
+    ));
+    ResponsesWebSocketConnection {
+        connection_id,
+        opened_at,
+        expired,
+        commands: Some(command_tx),
+        incoming: incoming_rx,
+        deferred: VecDeque::new(),
+        exited: exit_rx,
+        pump_task: Some(pump_task),
+        stats,
+        config,
+        exit_reason: None,
+    }
+}

backend/crates/gateway-api/src/openai/responses/websocket/connection/middleware.rs

@@ -1,11 +1,13 @@
 //! 插件消息组合在 transport 锁之外执行，主动发送复用唯一 writer
-use super::{ConnectionWriteError, PumpContext, write_message};
+use super::{ConnectionWriteError, frame::DOWNSTREAM_CLOSE_TIMEOUT, state::PumpContext};
 pub(super) use crate::middleware::websocket::CancelOnDrop;
 use crate::middleware::websocket::{from_core, into_core};
 use axum::extract::ws::Message;
-use futures::{Sink, future::BoxFuture};
-use gateway_core::{engine::middleware::MiddlewareError, middleware::websocket as core};
-use std::sync::Arc;
+use futures::{Sink, SinkExt, future::BoxFuture};
+use gateway_core::engine::middleware::{MiddlewareError, websocket as core};
+use gateway_core::lifecycle::CancellationToken;
+use std::{fmt, sync::Arc};
+use tokio::time::timeout;
 
 pub(super) struct Writer<S> {
     socket: tokio::sync::Mutex<S>,
@@ -37,6 +39,36 @@ where
             }) => result.unwrap_or(Err(ConnectionWriteError::Timeout { timeout: self.context.config.write_timeout })),
         }
     }
+
+    pub(super) async fn finish_peer_close(
+        &self,
+        cancellation: &CancellationToken,
+    ) -> Result<(), ConnectionWriteError> {
+        // 主动 Sender 也可能在 pump 之外持锁；先独占 transport 再取消业务，
+        // 有在途写或已取消的半帧时立即跳过，不能等待它释放锁后继续刷新
+        let socket = self
+            .socket
+            .try_lock()
+            .ok()
+            .filter(|_| !self.context.cancellation.is_cancelled());
+        self.context.cancellation.cancel();
+        let Some(mut socket) = socket else {
+            return Err(ConnectionWriteError::Closed);
+        };
+        // 收到 Close 后 transport 已排队应答；只刷新它，不能再发送一份 Close
+        // 业务调用已取消，清理只受宿主取消与独立短预算约束，不沿用业务写入的长超时
+        tokio::select! {
+            biased;
+            () = cancellation.cancelled() => Err(ConnectionWriteError::Closed),
+            result = timeout(DOWNSTREAM_CLOSE_TIMEOUT, async {
+                socket.flush().await.map_err(|error| {
+                    ConnectionWriteError::Transport { message: error.to_string() }
+                })
+            }) => result.unwrap_or(Err(ConnectionWriteError::Timeout {
+                timeout: DOWNSTREAM_CLOSE_TIMEOUT,
+            })),
+        }
+    }
 }
 impl<S, E> core::Sender for Writer<S>
 where
@@ -81,3 +113,25 @@ pub(super) async fn transform(
     .map(from_core)
     .transpose()
 }
+
+async fn write_message<S, E>(
+    socket: &mut S,
+    message: Message,
+    context: &PumpContext,
+) -> Result<(), ConnectionWriteError>
+where
+    S: Sink<Message, Error = E> + Unpin,
+    E: fmt::Display,
+{
+    tokio::select! {
+        biased;
+        () = context.cancellation.cancelled() => Err(ConnectionWriteError::Closed),
+        result = timeout(context.config.write_timeout, socket.send(message)) => {
+            match result {
+                Ok(Ok(())) => { context.stats.record_write(context.opened_at); Ok(()) }
+                Ok(Err(error)) => Err(ConnectionWriteError::Transport { message: error.to_string() }),
+                Err(_) => Err(ConnectionWriteError::Timeout { timeout: context.config.write_timeout }),
+            }
+        }
+    }
+}

backend/crates/gateway-api/src/openai/responses/websocket/connection/mod.rs

@@ -1,826 +1,13 @@
-//! 下游客户端 WebSocket 的单 owner pump 与有界收发边界
+//! 下游客户端 WebSocket 的连接合同、句柄与单 owner pump
 
+mod frame;
+mod handle;
 mod middleware;
+mod pump;
+mod state;
 
-use std::{
-    collections::VecDeque,
-    fmt,
-    sync::{
-        Arc,
-        atomic::{AtomicBool, AtomicU64, AtomicUsize, Ordering},
-    },
-    time::Duration,
+pub use frame::{
+    ConnectionConfig, ConnectionEvent, ConnectionWriteError, FramePhase, PumpExitReason,
+    WriteContext, WriteOutcome,
 };
-
-use axum::extract::ws::{CloseFrame, Message, WebSocket, close_code};
-use futures::{Sink, SinkExt, Stream, StreamExt};
-use gateway_core::{
-    engine::middleware::{FrozenMiddlewarePlan, MiddlewareHeader},
-    lifecycle::CancellationToken,
-    middleware::websocket as contract,
-};
-use thiserror::Error;
-use tokio::{
-    sync::{OwnedSemaphorePermit, Semaphore, mpsc, oneshot},
-    task::JoinHandle,
-    time::{Instant, timeout},
-};
-
-const OUTBOUND_COMMAND_BUFFER: usize = 32;
-const INBOUND_EVENT_BUFFER: usize = 32;
-const DOWNSTREAM_WRITE_TIMEOUT: Duration = Duration::from_secs(5 * 60);
-const CONNECTION_MAX_AGE: Duration = Duration::from_secs(60 * 60);
-const DOWNSTREAM_PING_INTERVAL: Duration = Duration::from_secs(25);
-
-/// WebSocket pump 的写入与生命周期预算
-#[derive(Clone, Copy)]
-pub struct ConnectionConfig {
-    write_timeout: Duration,
-    max_age: Duration,
-}
-
-impl ConnectionConfig {
-    /// 生产环境固定预算
-    pub const PRODUCTION: Self = Self {
-        write_timeout: DOWNSTREAM_WRITE_TIMEOUT,
-        max_age: CONNECTION_MAX_AGE,
-    };
-}
-
-/// 业务层可观察的客户端输入；Ping/Pong 始终由 pump 消费
-pub enum ConnectionEvent {
-    Text(String),
-    Binary,
-    Expired,
-    Exited(PumpExitReason),
-}
-
-/// 接收队列与活动期间暂存的请求共用容量；移动帧不释放它占用的名额
-pub(super) struct PendingConnectionEvent {
-    pub(super) event: ConnectionEvent,
-    _permit: Option<OwnedSemaphorePermit>,
-}
-
-impl PendingConnectionEvent {
-    fn exited(reason: PumpExitReason) -> Self {
-        Self {
-            event: ConnectionEvent::Exited(reason),
-            _permit: None,
-        }
-    }
-}
-
-/// 下游写入阶段；名称刻意使用 write，而不是暗示客户端已消费的 delivery
-#[derive(Debug, Clone, Copy, PartialEq, Eq)]
-pub enum FramePhase {
-    Metadata,
-    First,
-    Data,
-    Terminal,
-    FirstAndTerminal,
-    Error,
-    ConnectionLimit,
-    Close,
-}
-
-impl FramePhase {
-    const fn as_str(self) -> &'static str {
-        match self {
-            Self::Metadata => "metadata",
-            Self::First => "first",
-            Self::Data => "data",
-            Self::Terminal => "terminal",
-            Self::FirstAndTerminal => "first_and_terminal",
-            Self::Error => "error",
-            Self::ConnectionLimit => "connection_limit",
-            Self::Close => "close",
-        }
-    }
-
-    const fn is_milestone(self) -> bool {
-        matches!(
-            self,
-            Self::First
-                | Self::Terminal
-                | Self::FirstAndTerminal
-                | Self::Error
-                | Self::ConnectionLimit
-                | Self::Close
-        )
-    }
-}
-
-/// 一次下游写入的请求归属与协议阶段
-#[derive(Clone)]
-pub struct WriteContext {
-    request_id: Option<Arc<str>>,
-    phase: FramePhase,
-}
-
-impl WriteContext {
-    /// 创建归属于某个请求的写入上下文
-    #[must_use]
-    pub fn request(request_id: &Arc<str>, phase: FramePhase) -> Self {
-        Self {
-            request_id: Some(Arc::clone(request_id)),
-            phase,
-        }
-    }
-
-    /// 创建连接级写入上下文
-    #[must_use]
-    pub const fn connection(phase: FramePhase) -> Self {
-        Self {
-            request_id: None,
-            phase,
-        }
-    }
-}
-
-/// WebSocket pump 停止的稳定原因
-#[derive(Debug, Clone, Copy, PartialEq, Eq)]
-pub enum PumpExitReason {
-    ClientClose,
-    PeerEof,
-    ReadError,
-    WriteError,
-    WriteTimeout,
-    LifecycleShutdown,
-    ConnectionMaxAge,
-    CoordinatorDropped,
-    InboundOverload,
-    ServerClose,
-    PumpStopped,
-    MiddlewareFailed,
-}
-
-impl PumpExitReason {
-    pub(super) const fn as_str(self) -> &'static str {
-        match self {
-            Self::ClientClose => "client_close",
-            Self::PeerEof => "peer_eof",
-            Self::ReadError => "read_error",
-            Self::WriteError => "write_error",
-            Self::WriteTimeout => "write_timeout",
-            Self::LifecycleShutdown => "lifecycle_shutdown",
-            Self::ConnectionMaxAge => "connection_max_age",
-            Self::CoordinatorDropped => "coordinator_dropped",
-            Self::InboundOverload => "inbound_overload",
-            Self::ServerClose => "server_close",
-            Self::PumpStopped => "pump_stopped",
-            Self::MiddlewareFailed => "middleware_failed",
-        }
-    }
-}
-
-/// 下游 WebSocket 写入失败
-#[derive(Debug, Error)]
-pub enum ConnectionWriteError {
-    #[error("downstream WebSocket pump is closed")]
-    Closed,
-    #[error("downstream WebSocket write timed out after {timeout:?}")]
-    Timeout { timeout: Duration },
-    #[error("downstream WebSocket transport write failed: {message}")]
-    Transport { message: String },
-}
-
-/// 只有实际 transport 写入才记为 Written；插件丢弃消息不伪造写入成功
-#[derive(Clone, Copy, Debug, PartialEq, Eq)]
-pub enum WriteOutcome {
-    Written,
-    Suppressed,
-}
-
-struct ConnectionCommand {
-    message: Message,
-    context: WriteContext,
-    acknowledged: oneshot::Sender<Result<WriteOutcome, ConnectionWriteError>>,
-}
-
-#[derive(Default)]
-struct ConnectionStats {
-    command_queue_high_water: AtomicUsize,
-    command_backpressure_count: AtomicU64,
-    ping_received_count: AtomicU64,
-    ping_written_count: AtomicU64,
-    pong_received_count: AtomicU64,
-    last_read_ms: AtomicU64,
-    last_write_ms: AtomicU64,
-}
-
-impl ConnectionStats {
-    fn record_read(&self, opened_at: Instant) {
-        self.last_read_ms.store(
-            u64::try_from(opened_at.elapsed().as_millis()).unwrap_or(u64::MAX),
-            Ordering::Relaxed,
-        );
-    }
-
-    fn record_write(&self, opened_at: Instant) {
-        self.last_write_ms.store(
-            u64::try_from(opened_at.elapsed().as_millis()).unwrap_or(u64::MAX),
-            Ordering::Relaxed,
-        );
-    }
-
-    fn observe_command_queue(&self, sender: &mpsc::Sender<ConnectionCommand>) {
-        let queued = OUTBOUND_COMMAND_BUFFER
-            .saturating_sub(sender.capacity())
-            .saturating_add(1)
-            .min(OUTBOUND_COMMAND_BUFFER);
-        self.command_queue_high_water
-            .fetch_max(queued, Ordering::Relaxed);
-        if sender.capacity() == 0 {
-            self.command_backpressure_count
-                .fetch_add(1, Ordering::Relaxed);
-        }
-    }
-}
-
-/// 协调层持有的单 owner WebSocket 连接句柄
-pub struct ResponsesWebSocketConnection {
-    connection_id: Arc<str>,
-    opened_at: Instant,
-    expired: Arc<AtomicBool>,
-    commands: Option<mpsc::Sender<ConnectionCommand>>,
-    incoming: mpsc::Receiver<PendingConnectionEvent>,
-    deferred: VecDeque<PendingConnectionEvent>,
-    exited: oneshot::Receiver<PumpExitReason>,
-    pump_task: Option<JoinHandle<()>>,
-    stats: Arc<ConnectionStats>,
-    config: ConnectionConfig,
-    exit_reason: Option<PumpExitReason>,
-}
-
-impl ResponsesWebSocketConnection {
-    pub(super) fn new(
-        socket: WebSocket,
-        connection_id: String,
-        cancellation: CancellationToken,
-        middleware: Option<FrozenMiddlewarePlan>,
-        headers: Arc<[MiddlewareHeader]>,
-    ) -> Self {
-        spawn_with_middleware(
-            socket,
-            Arc::<str>::from(connection_id),
-            cancellation,
-            ConnectionConfig::PRODUCTION,
-            middleware,
-            headers,
-        )
-    }
-
-    pub(super) fn id(&self) -> &str {
-        &self.connection_id
-    }
-
-    /// 返回连接是否已经达到生命周期上限
-    #[must_use]
-    pub fn is_expired(&self) -> bool {
-        self.expired.load(Ordering::Acquire)
-    }
-
-    pub(super) fn age(&self) -> Duration {
-        self.opened_at.elapsed()
-    }
-
-    /// 等待下一个需要业务层处理的客户端事件
-    pub async fn next_event(&mut self) -> Option<ConnectionEvent> {
-        if self.exit_reason.is_none() {
-            match self.exited.try_recv() {
-                Ok(reason) => {
-                    self.exit_reason = Some(reason);
-                }
-                Err(tokio::sync::oneshot::error::TryRecvError::Closed) => {
-                    self.exit_reason = Some(PumpExitReason::PumpStopped);
-                }
-                Err(tokio::sync::oneshot::error::TryRecvError::Empty) => {}
-            }
-        }
-        if let Some(reason) = self.exit_reason {
-            return Some(ConnectionEvent::Exited(reason));
-        }
-        if let Some(event) = self.deferred.pop_front() {
-            return Some(event.event);
-        }
-        self.next_active_event().await.map(|event| event.event)
-    }
-
-    /// 活动响应期间只消费新到的帧，已排队的下一轮请求继续保持串行
-    pub(super) async fn next_active_event(&mut self) -> Option<PendingConnectionEvent> {
-        if let Some(reason) = self.exit_reason {
-            return Some(PendingConnectionEvent::exited(reason));
-        }
-        let event = tokio::select! {
-            biased;
-            reason = &mut self.exited => {
-                Some(PendingConnectionEvent::exited(reason.unwrap_or(PumpExitReason::PumpStopped)))
-            }
-            event = self.incoming.recv() => event,
-        };
-        if let Some(PendingConnectionEvent {
-            event: ConnectionEvent::Exited(reason),
-            ..
-        }) = event.as_ref()
-        {
-            self.exit_reason.get_or_insert(*reason);
-        } else if event.is_none() {
-            self.exit_reason.get_or_insert(PumpExitReason::PumpStopped);
-        }
-        event
-    }
-
-    pub(super) fn defer(&mut self, event: PendingConnectionEvent) {
-        self.deferred.push_back(event);
-    }
-
-    /// 等待连接退出，不消费留给后续串行请求的业务帧
-    pub async fn wait_for_exit(&mut self) -> PumpExitReason {
-        if let Some(reason) = self.exit_reason {
-            return reason;
-        }
-        let reason = (&mut self.exited)
-            .await
-            .unwrap_or(PumpExitReason::PumpStopped);
-        self.exit_reason = Some(reason);
-        reason
-    }
-
-    /// 串行写入文本帧，并等待 pump 确认 transport 写入结果
-    ///
-    /// # Errors
-    ///
-    /// pump 已关闭、写入超时或 transport 失败时返回稳定错误
-    pub async fn send_text(
-        &mut self,
-        payload: String,
-        context: WriteContext,
-    ) -> Result<WriteOutcome, ConnectionWriteError> {
-        self.send(Message::Text(payload.into()), context).await
-    }
-
-    pub(super) async fn close_policy(
-        &mut self,
-        reason: &'static str,
-        request_id: Option<&Arc<str>>,
-    ) {
-        let context = request_id.map_or_else(
-            || WriteContext::connection(FramePhase::Close),
-            |request_id| WriteContext::request(request_id, FramePhase::Close),
-        );
-        self.close(
-            CloseFrame {
-                code: close_code::POLICY,
-                reason: reason.into(),
-            },
-            context,
-            PumpExitReason::ServerClose,
-        )
-        .await;
-    }
-
-    pub(super) async fn close_for_connection_limit(&mut self, reason: &'static str) {
-        self.close(
-            CloseFrame {
-                code: close_code::NORMAL,
-                reason: reason.into(),
-            },
-            WriteContext::connection(FramePhase::Close),
-            PumpExitReason::ConnectionMaxAge,
-        )
-        .await;
-    }
-
-    async fn close(
-        &mut self,
-        frame: CloseFrame,
-        context: WriteContext,
-        exit_reason: PumpExitReason,
-    ) {
-        if self
-            .send(Message::Close(Some(frame)), context)
-            .await
-            .is_ok()
-        {
-            self.exit_reason.get_or_insert(exit_reason);
-        }
-    }
-
-    async fn send(
-        &mut self,
-        message: Message,
-        context: WriteContext,
-    ) -> Result<WriteOutcome, ConnectionWriteError> {
-        let Some(commands) = self.commands.as_ref().cloned() else {
-            return Err(ConnectionWriteError::Closed);
-        };
-        self.stats.observe_command_queue(&commands);
-        let (acknowledged, acknowledgement) = oneshot::channel();
-        let started_at = Instant::now();
-        let request_id = context.request_id.clone();
-        let phase = context.phase;
-        let command = ConnectionCommand {
-            message,
-            context,
-            acknowledged,
-        };
-        let write = async move {
-            commands
-                .send(command)
-                .await
-                .map_err(|_| ConnectionWriteError::Closed)?;
-            acknowledgement
-                .await
-                .map_err(|_| ConnectionWriteError::Closed)?
-        };
-        let result = match timeout(self.config.write_timeout, write).await {
-            Ok(result) => result,
-            Err(_) => Err(ConnectionWriteError::Timeout {
-                timeout: self.config.write_timeout,
-            }),
-        };
-        let duration = started_at.elapsed();
-        match &result {
-            Ok(WriteOutcome::Written) if phase.is_milestone() => tracing::info!(
-                websocket_connection_id = %self.connection_id,
-                request_id = request_id.as_deref().unwrap_or(""),
-                frame_phase = phase.as_str(),
-                write_duration_ms = duration.as_millis(),
-                "Responses WebSocket frame write succeeded"
-            ),
-            Ok(outcome) => tracing::debug!(
-                websocket_connection_id = %self.connection_id,
-                request_id = request_id.as_deref().unwrap_or(""),
-                frame_phase = phase.as_str(),
-                write_duration_ms = duration.as_millis(),
-                ?outcome,
-                "Responses WebSocket frame processed"
-            ),
-            Err(error) => tracing::info!(
-                websocket_connection_id = %self.connection_id,
-                request_id = request_id.as_deref().unwrap_or(""),
-                frame_phase = phase.as_str(),
-                write_duration_ms = duration.as_millis(),
-                error = %error,
-                "Responses WebSocket frame write failed"
-            ),
-        }
-        if let Err(error) = &result {
-            let reason = match error {
-                ConnectionWriteError::Timeout { .. } => PumpExitReason::WriteTimeout,
-                ConnectionWriteError::Transport { .. } => PumpExitReason::WriteError,
-                ConnectionWriteError::Closed => PumpExitReason::PumpStopped,
-            };
-            self.terminate(reason);
-        }
-        result
-    }
-
-    fn terminate(&mut self, reason: PumpExitReason) {
-        self.exit_reason.get_or_insert(reason);
-        self.commands.take();
-        if let Some(task) = self.pump_task.take() {
-            task.abort();
-        }
-    }
-
-    pub(super) fn log_summary(&self, request_count: u64) {
-        let connection_age_ms = self.age().as_millis();
-        let read_idle_ms = connection_age_ms
-            .saturating_sub(u128::from(self.stats.last_read_ms.load(Ordering::Relaxed)));
-        let write_idle_ms = connection_age_ms
-            .saturating_sub(u128::from(self.stats.last_write_ms.load(Ordering::Relaxed)));
-        tracing::info!(
-            websocket_connection_id = %self.connection_id,
-            request_count,
-            connection_age_ms,
-            pump_exit_reason = self
-                .exit_reason
-                .map_or("unknown", PumpExitReason::as_str),
-            command_queue_high_water = self
-                .stats
-                .command_queue_high_water
-                .load(Ordering::Relaxed),
-            command_backpressure_count = self
-                .stats
-                .command_backpressure_count
-                .load(Ordering::Relaxed),
-            ping_received_count = self.stats.ping_received_count.load(Ordering::Relaxed),
-            ping_written_count = self.stats.ping_written_count.load(Ordering::Relaxed),
-            pong_received_count = self.stats.pong_received_count.load(Ordering::Relaxed),
-            read_idle_ms,
-            write_idle_ms,
-            "Responses WebSocket disconnected"
-        );
-    }
-}
-
-impl Drop for ResponsesWebSocketConnection {
-    fn drop(&mut self) {
-        self.commands.take();
-        if let Some(task) = self.pump_task.take() {
-            task.abort();
-        }
-    }
-}
-
-/// 为自动处理 Ping/Pong 的 WebSocket transport 启动单 owner pump
-pub fn spawn_connection<S, E>(
-    socket: S,
-    connection_id: Arc<str>,
-    cancellation: CancellationToken,
-    config: ConnectionConfig,
-) -> ResponsesWebSocketConnection
-where
-    S: Stream<Item = Result<Message, E>> + Sink<Message, Error = E> + Unpin + Send + 'static,
-    E: fmt::Display + Send + 'static,
-{
-    spawn_with_middleware(
-        socket,
-        connection_id,
-        cancellation,
-        config,
-        None,
-        Arc::from([]),
-    )
-}
-
-fn spawn_with_middleware<S, E>(
-    socket: S,
-    connection_id: Arc<str>,
-    cancellation: CancellationToken,
-    config: ConnectionConfig,
-    middleware: Option<FrozenMiddlewarePlan>,
-    headers: Arc<[MiddlewareHeader]>,
-) -> ResponsesWebSocketConnection
-where
-    S: Stream<Item = Result<Message, E>> + Sink<Message, Error = E> + Unpin + Send + 'static,
-    E: fmt::Display + Send + 'static,
-{
-    let opened_at = Instant::now();
-    let expired = Arc::new(AtomicBool::new(false));
-    let stats = Arc::new(ConnectionStats::default());
-    let (command_tx, command_rx) = mpsc::channel(OUTBOUND_COMMAND_BUFFER);
-    let (incoming_tx, incoming_rx) = mpsc::channel(INBOUND_EVENT_BUFFER);
-    let (exit_tx, exit_rx) = oneshot::channel();
-    let pump_task = tokio::spawn(run_pump(
-        socket,
-        command_rx,
-        incoming_tx,
-        exit_tx,
-        Arc::clone(&connection_id),
-        cancellation,
-        opened_at,
-        Arc::clone(&expired),
-        Arc::clone(&stats),
-        config,
-        middleware,
-        headers,
-    ));
-    ResponsesWebSocketConnection {
-        connection_id,
-        opened_at,
-        expired,
-        commands: Some(command_tx),
-        incoming: incoming_rx,
-        deferred: VecDeque::new(),
-        exited: exit_rx,
-        pump_task: Some(pump_task),
-        stats,
-        config,
-        exit_reason: None,
-    }
-}
-
-#[expect(clippy::too_many_arguments)]
-async fn run_pump<S, E>(
-    socket: S,
-    commands: mpsc::Receiver<ConnectionCommand>,
-    incoming: mpsc::Sender<PendingConnectionEvent>,
-    exited: oneshot::Sender<PumpExitReason>,
-    connection_id: Arc<str>,
-    cancellation: CancellationToken,
-    opened_at: Instant,
-    expired: Arc<AtomicBool>,
-    stats: Arc<ConnectionStats>,
-    config: ConnectionConfig,
-    middleware: Option<FrozenMiddlewarePlan>,
-    headers: Arc<[MiddlewareHeader]>,
-) where
-    S: Stream<Item = Result<Message, E>> + Sink<Message, Error = E> + Unpin + Send + 'static,
-    E: fmt::Display + Send + 'static,
-{
-    let context = Arc::new(PumpContext {
-        connection_id,
-        cancellation: cancellation.child_token(),
-        opened_at,
-        expired,
-        stats,
-        config,
-        middleware,
-        headers,
-    });
-    let _lifetime = middleware::CancelOnDrop(Some(context.cancellation.clone()));
-    let (writer, mut reader) = socket.split();
-    let writer = Arc::new(middleware::Writer::new(writer, context.clone()));
-    let sender: Arc<dyn contract::Sender> = writer.clone();
-    // 同一个受管任务同时驱动两侧；等待写入不能阻塞入站控制、关闭或取消
-    let reason = tokio::select! {
-        biased;
-        reason = read_connection(&mut reader, &incoming, sender.clone(), &context) => reason,
-        reason = write_connection(writer.clone(), commands, &context) => reason,
-    };
-    context.cancellation.cancel();
-    let _ = exited.send(reason);
-    tracing::debug!(websocket_connection_id = %context.connection_id, pump_exit_reason = reason.as_str(), "Responses WebSocket pump exited");
-}
-
-struct PumpContext {
-    middleware: Option<FrozenMiddlewarePlan>,
-    headers: Arc<[MiddlewareHeader]>,
-    connection_id: Arc<str>,
-    cancellation: CancellationToken,
-    opened_at: Instant,
-    expired: Arc<AtomicBool>,
-    stats: Arc<ConnectionStats>,
-    config: ConnectionConfig,
-}
-
-async fn read_connection<S, E>(
-    socket: &mut S,
-    incoming: &mpsc::Sender<PendingConnectionEvent>,
-    sender: Arc<dyn contract::Sender>,
-    context: &PumpContext,
-) -> PumpExitReason
-where
-    S: Stream<Item = Result<Message, E>> + Unpin,
-    E: fmt::Display,
-{
-    let event_slots = Arc::new(Semaphore::new(INBOUND_EVENT_BUFFER));
-    let deadline = tokio::time::sleep_until(context.opened_at + context.config.max_age);
-    tokio::pin!(deadline);
-    let mut deadline_elapsed = false;
-    loop {
-        tokio::select! {
-            biased;
-            () = context.cancellation.cancelled() => return PumpExitReason::LifecycleShutdown,
-            () = &mut deadline, if !deadline_elapsed => {
-                deadline_elapsed = true;
-                context.expired.store(true, Ordering::Release);
-                // 到期只阻止下一轮，当前响应仍由协调层按原有合同收尾
-                match emit_incoming(incoming, &event_slots, ConnectionEvent::Expired) {
-                    Ok(()) | Err(PumpExitReason::InboundOverload) => {}
-                    Err(reason) => return reason,
-                }
-            }
-            message = socket.next() => {
-                if matches!(&message, Some(Ok(_))) { context.stats.record_read(context.opened_at); }
-                let message = match message {
-                    Some(Ok(message)) => match middleware::transform(message, contract::Direction::Incoming, sender.clone(), context).await {
-                        Ok(Some(message)) => Some(Ok(message)),
-                        Ok(None) => continue,
-                        Err(_) => return PumpExitReason::MiddlewareFailed,
-                    },
-                    other => other,
-                };
-                let event = match message {
-                    Some(Ok(Message::Text(payload))) => Some(ConnectionEvent::Text(payload.to_string())),
-                    Some(Ok(Message::Binary(_))) => Some(ConnectionEvent::Binary),
-                    Some(Ok(Message::Ping(_))) => {
-                        // Axum/tungstenite 在继续读取时自动刷新 Pong，不能再手工发送一份
-                        context.stats.ping_received_count.fetch_add(1, Ordering::Relaxed);
-                        None
-                    }
-                    Some(Ok(Message::Pong(_))) => {
-                        context.stats.pong_received_count.fetch_add(1, Ordering::Relaxed);
-                        None
-                    }
-                    Some(Ok(Message::Close(_))) => return PumpExitReason::ClientClose,
-                    Some(Err(error)) => {
-                        tracing::info!(websocket_connection_id = %context.connection_id, error = %error, "Responses WebSocket receive failed");
-                        return PumpExitReason::ReadError;
-                    }
-                    None => return PumpExitReason::PeerEof,
-                };
-                if let Some(event) = event && let Err(reason) = emit_incoming(incoming, &event_slots, event) { return reason; }
-            }
-        }
-    }
-}
-
-async fn write_connection<S, E>(
-    writer: Arc<middleware::Writer<S>>,
-    mut commands: mpsc::Receiver<ConnectionCommand>,
-    context: &PumpContext,
-) -> PumpExitReason
-where
-    S: Sink<Message, Error = E> + Unpin + Send + 'static,
-    E: fmt::Display + Send + 'static,
-{
-    let mut heartbeat = tokio::time::interval_at(
-        context.opened_at + DOWNSTREAM_PING_INTERVAL,
-        DOWNSTREAM_PING_INTERVAL,
-    );
-    heartbeat.set_missed_tick_behavior(tokio::time::MissedTickBehavior::Skip);
-    let mut ping_sequence = 0_u64;
-    loop {
-        let message = tokio::select! {
-            biased;
-            () = context.cancellation.cancelled() => return PumpExitReason::LifecycleShutdown,
-            command = commands.recv() => {
-                let Some(command) = command else { return PumpExitReason::CoordinatorDropped; };
-                let ConnectionCommand { message, context: write_context, acknowledged } = command;
-                let closing = matches!(message, Message::Close(_));
-                let result = match middleware::transform(message, contract::Direction::Outgoing, writer.clone(), context).await {
-                    Ok(Some(message)) => writer.write(message).await.map(|()| WriteOutcome::Written),
-                    Ok(None) => Ok(WriteOutcome::Suppressed),
-                    Err(error) => Err(ConnectionWriteError::Transport { message: error.to_string() }),
-                };
-                let reason = result.as_ref().err().map(|error| write_exit_reason(error, context));
-                if let Err(error) = &result {
-                    tracing::debug!(websocket_connection_id = %context.connection_id, request_id = write_context.request_id.as_deref().unwrap_or(""), frame_phase = write_context.phase.as_str(), error = %error, "Responses WebSocket pump transport write failed");
-                }
-                let _ = acknowledged.send(result);
-                if let Some(reason) = reason { return reason; }
-                if closing { return PumpExitReason::ServerClose; }
-                continue;
-            }
-            _ = heartbeat.tick() => {
-                ping_sequence = ping_sequence.wrapping_add(1);
-                Message::Ping(ping_sequence.to_be_bytes().to_vec().into())
-            }
-        };
-        let message = match middleware::transform(
-            message,
-            contract::Direction::Outgoing,
-            writer.clone(),
-            context,
-        )
-        .await
-        {
-            Ok(Some(message)) => message,
-            Ok(None) => continue,
-            Err(_) => return PumpExitReason::MiddlewareFailed,
-        };
-        if let Err(error) = writer.write(message).await {
-            tracing::info!(websocket_connection_id = %context.connection_id, error = %error, "Responses WebSocket control write failed");
-            return write_exit_reason(&error, context);
-        }
-        context
-            .stats
-            .ping_written_count
-            .fetch_add(1, Ordering::Relaxed);
-    }
-}
-
-async fn write_message<S, E>(
-    socket: &mut S,
-    message: Message,
-    context: &PumpContext,
-) -> Result<(), ConnectionWriteError>
-where
-    S: Sink<Message, Error = E> + Unpin,
-    E: fmt::Display,
-{
-    tokio::select! {
-        biased;
-        () = context.cancellation.cancelled() => Err(ConnectionWriteError::Closed),
-        result = timeout(context.config.write_timeout, socket.send(message)) => {
-            match result {
-                Ok(Ok(())) => { context.stats.record_write(context.opened_at); Ok(()) }
-                Ok(Err(error)) => Err(ConnectionWriteError::Transport { message: error.to_string() }),
-                Err(_) => Err(ConnectionWriteError::Timeout { timeout: context.config.write_timeout }),
-            }
-        }
-    }
-}
-
-fn write_exit_reason(error: &ConnectionWriteError, context: &PumpContext) -> PumpExitReason {
-    match error {
-        ConnectionWriteError::Timeout { .. } => PumpExitReason::WriteTimeout,
-        ConnectionWriteError::Transport { .. } => PumpExitReason::WriteError,
-        ConnectionWriteError::Closed if context.cancellation.is_cancelled() => {
-            PumpExitReason::LifecycleShutdown
-        }
-        ConnectionWriteError::Closed => PumpExitReason::PumpStopped,
-    }
-}
-
-fn emit_incoming(
-    incoming: &mpsc::Sender<PendingConnectionEvent>,
-    slots: &Arc<Semaphore>,
-    event: ConnectionEvent,
-) -> Result<(), PumpExitReason> {
-    let permit = Arc::clone(slots)
-        .try_acquire_owned()
-        .map_err(|_| PumpExitReason::InboundOverload)?;
-    incoming
-        .try_send(PendingConnectionEvent {
-            event,
-            _permit: Some(permit),
-        })
-        .map_err(|error| match error {
-            mpsc::error::TrySendError::Full(_) => PumpExitReason::InboundOverload,
-            mpsc::error::TrySendError::Closed(_) => PumpExitReason::CoordinatorDropped,
-        })
-}
+pub use handle::{ResponsesWebSocketConnection, spawn_connection};

backend/crates/gateway-api/src/openai/responses/websocket/connection/pump.rs

@@ -0,0 +1,230 @@
+//! 单 owner 驱动连接读写、心跳与有界入站队列
+
+use super::{
+    frame::*,
+    middleware,
+    state::{ConnectionCommand, ConnectionStats, PumpContext},
+};
+use axum::extract::ws::Message;
+use futures::{Sink, Stream, StreamExt};
+use gateway_core::{
+    engine::middleware::{FrozenMiddlewarePlan, MiddlewareHeader, websocket as contract},
+    lifecycle::CancellationToken,
+};
+use std::{
+    fmt,
+    sync::{
+        Arc,
+        atomic::{AtomicBool, Ordering},
+    },
+};
+use tokio::{
+    sync::{Semaphore, mpsc, oneshot},
+    time::Instant,
+};
+
+#[expect(clippy::too_many_arguments)]
+pub(super) async fn run_pump<S, E>(
+    socket: S,
+    commands: mpsc::Receiver<ConnectionCommand>,
+    incoming: mpsc::Sender<PendingConnectionEvent>,
+    exited: oneshot::Sender<PumpExitReason>,
+    connection_id: Arc<str>,
+    cancellation: CancellationToken,
+    opened_at: Instant,
+    expired: Arc<AtomicBool>,
+    stats: Arc<ConnectionStats>,
+    config: ConnectionConfig,
+    middleware: Option<FrozenMiddlewarePlan>,
+    headers: Arc<[MiddlewareHeader]>,
+) where
+    S: Stream<Item = Result<Message, E>> + Sink<Message, Error = E> + Unpin + Send + 'static,
+    E: fmt::Display + Send + 'static,
+{
+    let context = Arc::new(PumpContext {
+        connection_id,
+        cancellation: cancellation.child_token(),
+        opened_at,
+        expired,
+        stats,
+        config,
+        middleware,
+        headers,
+    });
+    let _lifetime = middleware::CancelOnDrop(Some(context.cancellation.clone()));
+    let (writer, mut reader) = socket.split();
+    let writer = Arc::new(middleware::Writer::new(writer, context.clone()));
+    let sender: Arc<dyn contract::Sender> = writer.clone();
+    // 同一个受管任务同时驱动两侧；等待写入不能阻塞入站控制、关闭或取消
+    let reason = tokio::select! {
+        biased;
+        reason = read_connection(&mut reader, &incoming, sender.clone(), &context) => reason,
+        reason = write_connection(writer.clone(), commands, &context) => reason,
+    };
+    if matches!(reason, PumpExitReason::ClientClose) {
+        if let Err(error) = writer.finish_peer_close(&cancellation).await {
+            tracing::debug!(websocket_connection_id = %context.connection_id, %error, "Responses WebSocket close acknowledgement could not be flushed");
+        }
+    } else {
+        context.cancellation.cancel();
+    }
+    let _ = exited.send(reason);
+    tracing::debug!(websocket_connection_id = %context.connection_id, pump_exit_reason = reason.as_str(), "Responses WebSocket pump exited");
+}
+
+async fn read_connection<S, E>(
+    socket: &mut S,
+    incoming: &mpsc::Sender<PendingConnectionEvent>,
+    sender: Arc<dyn contract::Sender>,
+    context: &PumpContext,
+) -> PumpExitReason
+where
+    S: Stream<Item = Result<Message, E>> + Unpin,
+    E: fmt::Display,
+{
+    let event_slots = Arc::new(Semaphore::new(INBOUND_EVENT_BUFFER));
+    let deadline = tokio::time::sleep_until(context.opened_at + context.config.max_age);
+    tokio::pin!(deadline);
+    let mut deadline_elapsed = false;
+    loop {
+        tokio::select! {
+            biased;
+            () = context.cancellation.cancelled() => return PumpExitReason::LifecycleShutdown,
+            () = &mut deadline, if !deadline_elapsed => {
+                deadline_elapsed = true;
+                context.expired.store(true, Ordering::Release);
+                // 到期只阻止下一轮，当前响应仍由协调层按原有合同收尾
+                match emit_incoming(incoming, &event_slots, ConnectionEvent::Expired) {
+                    Ok(()) | Err(PumpExitReason::InboundOverload) => {}
+                    Err(reason) => return reason,
+                }
+            }
+            message = socket.next() => {
+                if matches!(&message, Some(Ok(_))) { context.stats.record_read(context.opened_at); }
+                let message = match message {
+                    Some(Ok(message)) => match middleware::transform(message, contract::Direction::Incoming, sender.clone(), context).await {
+                        Ok(Some(message)) => Some(Ok(message)),
+                        Ok(None) => continue,
+                        Err(_) => return PumpExitReason::MiddlewareFailed,
+                    },
+                    other => other,
+                };
+                let event = match message {
+                    Some(Ok(Message::Text(payload))) => Some(ConnectionEvent::Text(payload.to_string())),
+                    Some(Ok(Message::Binary(_))) => Some(ConnectionEvent::Binary),
+                    Some(Ok(Message::Ping(_))) => {
+                        // Axum/tungstenite 在继续读取时自动刷新 Pong，不能再手工发送一份
+                        context.stats.ping_received_count.fetch_add(1, Ordering::Relaxed);
+                        None
+                    }
+                    Some(Ok(Message::Pong(_))) => {
+                        context.stats.pong_received_count.fetch_add(1, Ordering::Relaxed);
+                        None
+                    }
+                    Some(Ok(Message::Close(_))) => return PumpExitReason::ClientClose,
+                    Some(Err(error)) => {
+                        tracing::info!(websocket_connection_id = %context.connection_id, error = %error, "Responses WebSocket receive failed");
+                        return PumpExitReason::ReadError;
+                    }
+                    None => return PumpExitReason::PeerEof,
+                };
+                if let Some(event) = event && let Err(reason) = emit_incoming(incoming, &event_slots, event) { return reason; }
+            }
+        }
+    }
+}
+
+async fn write_connection<S, E>(
+    writer: Arc<middleware::Writer<S>>,
+    mut commands: mpsc::Receiver<ConnectionCommand>,
+    context: &PumpContext,
+) -> PumpExitReason
+where
+    S: Sink<Message, Error = E> + Unpin + Send + 'static,
+    E: fmt::Display + Send + 'static,
+{
+    let mut heartbeat = tokio::time::interval_at(
+        context.opened_at + DOWNSTREAM_PING_INTERVAL,
+        DOWNSTREAM_PING_INTERVAL,
+    );
+    heartbeat.set_missed_tick_behavior(tokio::time::MissedTickBehavior::Skip);
+    let mut ping_sequence = 0_u64;
+    loop {
+        let message = tokio::select! {
+            biased;
+            () = context.cancellation.cancelled() => return PumpExitReason::LifecycleShutdown,
+            command = commands.recv() => {
+                let Some(command) = command else { return PumpExitReason::CoordinatorDropped; };
+                let ConnectionCommand { message, context: write_context, acknowledged } = command;
+                let closing = matches!(message, Message::Close(_));
+                let result = match middleware::transform(message, contract::Direction::Outgoing, writer.clone(), context).await {
+                    Ok(Some(message)) => writer.write(message).await.map(|()| WriteOutcome::Written),
+                    Ok(None) => Ok(WriteOutcome::Suppressed),
+                    Err(error) => Err(ConnectionWriteError::Transport { message: error.to_string() }),
+                };
+                let reason = result.as_ref().err().map(|error| write_exit_reason(error, context));
+                if let Err(error) = &result {
+                    tracing::debug!(websocket_connection_id = %context.connection_id, request_id = write_context.request_id.as_deref().unwrap_or(""), frame_phase = write_context.phase.as_str(), error = %error, "Responses WebSocket pump transport write failed");
+                }
+                let _ = acknowledged.send(result);
+                if let Some(reason) = reason { return reason; }
+                if closing { return PumpExitReason::ServerClose; }
+                continue;
+            }
+            _ = heartbeat.tick() => {
+                ping_sequence = ping_sequence.wrapping_add(1);
+                Message::Ping(ping_sequence.to_be_bytes().to_vec().into())
+            }
+        };
+        let message = match middleware::transform(
+            message,
+            contract::Direction::Outgoing,
+            writer.clone(),
+            context,
+        )
+        .await
+        {
+            Ok(Some(message)) => message,
+            Ok(None) => continue,
+            Err(_) => return PumpExitReason::MiddlewareFailed,
+        };
+        if let Err(error) = writer.write(message).await {
+            tracing::info!(websocket_connection_id = %context.connection_id, error = %error, "Responses WebSocket control write failed");
+            return write_exit_reason(&error, context);
+        }
+        context
+            .stats
+            .ping_written_count
+            .fetch_add(1, Ordering::Relaxed);
+    }
+}
+
+fn write_exit_reason(error: &ConnectionWriteError, context: &PumpContext) -> PumpExitReason {
+    match error {
+        ConnectionWriteError::Timeout { .. } => PumpExitReason::WriteTimeout,
+        ConnectionWriteError::Transport { .. } => PumpExitReason::WriteError,
+        ConnectionWriteError::Closed if context.cancellation.is_cancelled() => {
+            PumpExitReason::LifecycleShutdown
+        }
+        ConnectionWriteError::Closed => PumpExitReason::PumpStopped,
+    }
+}
+
+fn emit_incoming(
+    incoming: &mpsc::Sender<PendingConnectionEvent>,
+    slots: &Arc<Semaphore>,
+    event: ConnectionEvent,
+) -> Result<(), PumpExitReason> {
+    let permit = Arc::clone(slots)
+        .try_acquire_owned()
+        .map_err(|_| PumpExitReason::InboundOverload)?;
+    incoming
+        .try_send(PendingConnectionEvent {
+            event,
+            _permit: Some(permit),
+        })
+        .map_err(|error| match error {
+            mpsc::error::TrySendError::Full(_) => PumpExitReason::InboundOverload,
+            mpsc::error::TrySendError::Closed(_) => PumpExitReason::CoordinatorDropped,
+        })
+}

backend/crates/gateway-api/src/openai/responses/websocket/connection/state.rs

@@ -0,0 +1,75 @@
+//! 单连接共享的命令、运行预算与观测状态
+
+use super::frame::{
+    ConnectionConfig, ConnectionWriteError, OUTBOUND_COMMAND_BUFFER, WriteContext, WriteOutcome,
+};
+use axum::extract::ws::Message;
+use gateway_core::{
+    engine::middleware::{FrozenMiddlewarePlan, MiddlewareHeader},
+    lifecycle::CancellationToken,
+};
+use std::sync::{
+    Arc,
+    atomic::{AtomicBool, AtomicU64, AtomicUsize, Ordering},
+};
+use tokio::{
+    sync::{mpsc, oneshot},
+    time::Instant,
+};
+
+pub(super) struct ConnectionCommand {
+    pub(super) message: Message,
+    pub(super) context: WriteContext,
+    pub(super) acknowledged: oneshot::Sender<Result<WriteOutcome, ConnectionWriteError>>,
+}
+
+#[derive(Default)]
+pub(super) struct ConnectionStats {
+    pub(super) command_queue_high_water: AtomicUsize,
+    pub(super) command_backpressure_count: AtomicU64,
+    pub(super) ping_received_count: AtomicU64,
+    pub(super) ping_written_count: AtomicU64,
+    pub(super) pong_received_count: AtomicU64,
+    pub(super) last_read_ms: AtomicU64,
+    pub(super) last_write_ms: AtomicU64,
+}
+
+impl ConnectionStats {
+    pub(super) fn record_read(&self, opened_at: Instant) {
+        self.last_read_ms.store(
+            u64::try_from(opened_at.elapsed().as_millis()).unwrap_or(u64::MAX),
+            Ordering::Relaxed,
+        );
+    }
+
+    pub(super) fn record_write(&self, opened_at: Instant) {
+        self.last_write_ms.store(
+            u64::try_from(opened_at.elapsed().as_millis()).unwrap_or(u64::MAX),
+            Ordering::Relaxed,
+        );
+    }
+
+    pub(super) fn observe_command_queue(&self, sender: &mpsc::Sender<ConnectionCommand>) {
+        let queued = OUTBOUND_COMMAND_BUFFER
+            .saturating_sub(sender.capacity())
+            .saturating_add(1)
+            .min(OUTBOUND_COMMAND_BUFFER);
+        self.command_queue_high_water
+            .fetch_max(queued, Ordering::Relaxed);
+        if sender.capacity() == 0 {
+            self.command_backpressure_count
+                .fetch_add(1, Ordering::Relaxed);
+        }
+    }
+}
+
+pub(super) struct PumpContext {
+    pub(super) middleware: Option<FrozenMiddlewarePlan>,
+    pub(super) headers: Arc<[MiddlewareHeader]>,
+    pub(super) connection_id: Arc<str>,
+    pub(super) cancellation: CancellationToken,
+    pub(super) opened_at: Instant,
+    pub(super) expired: Arc<AtomicBool>,
+    pub(super) stats: Arc<ConnectionStats>,
+    pub(super) config: ConnectionConfig,
+}

backend/crates/gateway-api/tests/admin/account_groups.rs

@@ -49,6 +49,36 @@ async fn list_route_should_keep_camel_case_group_and_page_wire() {
     assert!(value["data"]["items"][0].get("member_count").is_none());
 }
 
+#[tokio::test]
+async fn options_route_should_return_only_lightweight_group_fields() {
+    let fixture = authenticated_fixture().await;
+    let response = request(
+        router(&fixture),
+        Method::GET,
+        "/api/admin/account-groups/options?page=1&pageSize=1&search=alpha&enabled=true",
+        None,
+        true,
+    )
+    .await;
+
+    assert_eq!(response.status(), StatusCode::OK);
+    let value = response_json(response).await;
+    assert_eq!(
+        value["data"]["items"][0],
+        json!({
+            "id": PRIMARY_GROUP_ID,
+            "name": "Alpha routing",
+            "color": "#2563EBFF",
+            "enabled": true
+        })
+    );
+    assert_eq!(value["data"]["page"]["page"], 1);
+    assert_eq!(value["data"]["page"]["pageSize"], 1);
+    assert_eq!(value["data"]["page"]["total"], 1);
+    assert_eq!(value["data"]["page"]["totalPages"], 1);
+    assert_eq!(value["data"]["configRevision"], 7);
+}
+
 #[tokio::test]
 async fn create_route_should_create_an_empty_group() {
     let fixture = authenticated_fixture().await;
@@ -163,6 +193,11 @@ async fn list_route_should_reject_unknown_and_invalid_pagination_fields() {
         "/api/admin/account-groups?page=0",
         "/api/admin/account-groups?pageSize=0",
         "/api/admin/account-groups?pageSize=201",
+        "/api/admin/account-groups/options?other=true",
+        "/api/admin/account-groups/options?page=0",
+        "/api/admin/account-groups/options?pageSize=0",
+        "/api/admin/account-groups/options?pageSize=201",
+        "/api/admin/account-groups/options?search=%00",
     ] {
         let response = request(router(&fixture), Method::GET, uri, None, true).await;
         assert_eq!(response.status(), StatusCode::BAD_REQUEST, "{uri}");
@@ -219,6 +254,14 @@ async fn account_group_routes_should_require_admin_authentication() {
         false,
     )
     .await;
+    let options_response = request(
+        router(&fixture),
+        Method::GET,
+        "/api/admin/account-groups/options",
+        None,
+        false,
+    )
+    .await;
     let post_response = request(
         router(&fixture),
         Method::POST,
@@ -229,9 +272,29 @@ async fn account_group_routes_should_require_admin_authentication() {
     .await;
 
     assert_eq!(get_response.status(), StatusCode::UNAUTHORIZED);
+    assert_eq!(options_response.status(), StatusCode::UNAUTHORIZED);
     assert_eq!(post_response.status(), StatusCode::UNAUTHORIZED);
 }
 
+#[tokio::test]
+async fn options_route_should_reject_client_key_identity() {
+    let fixture = crate::support::key_fixture().await;
+    let response = router(&fixture)
+        .oneshot(
+            Request::builder()
+                .method(Method::GET)
+                .uri("/api/admin/account-groups/options")
+                .header("x-api-key", crate::support::RAW_KEY)
+                .header("x-request-id", "req_account_group_options_client_key")
+                .body(Body::empty())
+                .expect("client key options request"),
+        )
+        .await
+        .expect("client key options response");
+
+    assert_eq!(response.status(), StatusCode::UNAUTHORIZED);
+}
+
 async fn authenticated_fixture() -> AdminTestFixture {
     let fixture = AdminTestFixture::new().await;
     fixture.auth.insert_session("valid-session");

backend/crates/gateway-api/tests/admin/mod.rs

@@ -18,9 +18,10 @@ use gateway_admin::{
         MutationContext, Revision,
         account_groups::{
             AccountGroupAccountSummary, AccountGroupCapacity, AccountGroupColor,
-            AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation, AccountGroupPage,
-            AccountGroupRecord, AccountGroupUsage, DeleteAccountGroup, NewAccountGroup,
-            SetAccountGroupEnabled, UpdateAccountGroup,
+            AccountGroupListQuery, AccountGroupMemberFact, AccountGroupMutation,
+            AccountGroupOptionsPage, AccountGroupPage, AccountGroupRecord, AccountGroupRef,
+            AccountGroupUsage, DeleteAccountGroup, NewAccountGroup, SetAccountGroupEnabled,
+            UpdateAccountGroup,
         },
         accounts::{
             AccountListQuery, AccountPage, AccountPageItem, AccountRuntimeSnapshot,
@@ -37,9 +38,10 @@ use gateway_admin::{
             NewClientKey, SetClientKeyEnabled, UpdateClientKey,
         },
         observability::{
-            DashboardObservation, DecimalAmount, DiagnosticDimension, DiagnosticsObservation,
-            OpsError, OpsErrorPage, OpsErrorQuery, RequestMetricPoint, TimeRange, UsageDetail,
-            UsageFilter, UsageListRecord, UsageOverview, UsagePage, UsageQuery,
+            DashboardObservation, DashboardQuery, DecimalAmount, DiagnosticDimension,
+            DiagnosticsObservation, Granularity, OpsError, OpsErrorPage, OpsErrorQuery,
+            RequestMetricPoint, TimeRange, UsageDetail, UsageFilter, UsageListRecord,
+            UsageOverview, UsagePage, UsageQuery,
         },
         provider_credentials::{
             AuthorizationCommit, AuthorizationStarted, CompleteAuthorization, CredentialDetails,
@@ -788,6 +790,48 @@ impl AccountGroupStore for MemoryAccountGroupStore {
         })
     }
 
+    async fn list_account_group_options(
+        &self,
+        query: AccountGroupListQuery,
+    ) -> AdminStoreResult<AccountGroupOptionsPage> {
+        let state = self.state.lock().expect("account groups");
+        let search = query.search.as_ref().map(|value| value.to_lowercase());
+        let mut matching = state
+            .groups
+            .values()
+            .filter(|record| {
+                query
+                    .enabled
+                    .is_none_or(|enabled| record.enabled == enabled)
+            })
+            .filter(|record| {
+                search
+                    .as_ref()
+                    .is_none_or(|search| record.name.to_lowercase().contains(search))
+            })
+            .map(|record| AccountGroupRef {
+                id: record.id.clone(),
+                name: record.name.clone(),
+                color: record.color.clone(),
+                enabled: record.enabled,
+            })
+            .collect::<Vec<_>>();
+        matching.sort_by(|left, right| left.name.cmp(&right.name));
+        let total = matching.len() as u64;
+        let page_size = usize::from(query.page_size.get());
+        let offset = usize::try_from(query.page.saturating_sub(1))
+            .unwrap_or(usize::MAX)
+            .saturating_mul(page_size);
+        let items = matching.into_iter().skip(offset).take(page_size).collect();
+        Ok(AccountGroupOptionsPage {
+            config_revision: state.revision,
+            items,
+            total,
+            page: query.page,
+            page_size: query.page_size.get(),
+        })
+    }
+
     async fn load_account_group_members(
         &self,
         group_ids: &[gateway_core::routing::AccountGroupId],
@@ -1303,13 +1347,13 @@ impl AccountRuntimeStore for UnusedStore {
 impl ObservabilityStore for UnusedStore {
     async fn dashboard_summary(
         &self,
-        range: TimeRange,
+        query: DashboardQuery,
         _: DateTime<Utc>,
     ) -> AdminStoreResult<DashboardObservation> {
         *self
             .dashboard_summary_range
             .lock()
-            .expect("dashboard summary range") = Some(range);
+            .expect("dashboard summary range") = Some(query.range);
         self.dashboard_observation
             .lock()
             .expect("dashboard observation")
@@ -1317,7 +1361,11 @@ impl ObservabilityStore for UnusedStore {
             .ok_or_else(|| unavailable("dashboard"))
     }
 
-    async fn dashboard_trend(&self, _: TimeRange) -> AdminStoreResult<Vec<RequestMetricPoint>> {
+    async fn dashboard_trend(
+        &self,
+        _: TimeRange,
+        _: Granularity,
+    ) -> AdminStoreResult<Vec<RequestMetricPoint>> {
         Err(unavailable("dashboard trend"))
     }
 
@@ -1325,6 +1373,7 @@ impl ObservabilityStore for UnusedStore {
         &self,
         range: TimeRange,
         filter: UsageFilter,
+        _: Granularity,
     ) -> AdminStoreResult<Vec<RequestMetricPoint>> {
         let mut data = self.observations.lock().expect("observations");
         data.trends.push((range, filter));
@@ -1335,6 +1384,7 @@ impl ObservabilityStore for UnusedStore {
         &self,
         _: TimeRange,
         _: UsageFilter,
+        _: Granularity,
     ) -> gateway_admin::ports::store::UsageCalculatedBillingStream<'_> {
         Box::pin(futures::stream::once(async {
             Err(unavailable("usage billing facts"))
@@ -1382,6 +1432,7 @@ impl ObservabilityStore for UnusedStore {
         _: TimeRange,
         _: UsageFilter,
         _: DiagnosticDimension,
+        _: u16,
     ) -> AdminStoreResult<DiagnosticsObservation> {
         Ok(self.diagnostics.lock().expect("diagnostics").clone())
     }

backend/crates/gateway-api/tests/admin/observability/response.rs

@@ -25,6 +25,7 @@ async fn dashboard_summary_should_include_quota_exhaustion_in_unavailable_headli
         .dashboard_observation
         .lock()
         .expect("dashboard observation") = Some(DashboardObservation {
+        runtime_slots: None,
         range: TimeRange::new(now - Duration::hours(1), now).expect("dashboard range"),
         totals: Default::default(),
         provider_accounts: AccountPoolMetrics {
@@ -107,6 +108,7 @@ async fn dashboard_summary_should_use_lifetime_totals_for_card_footers() {
         .dashboard_observation
         .lock()
         .expect("dashboard observation") = Some(DashboardObservation {
+        runtime_slots: None,
         range: TimeRange::new(now - Duration::hours(1), now).expect("dashboard range"),
         totals: DashboardTotals {
             request_count: 42,
@@ -178,6 +180,7 @@ async fn dashboard_summary_should_default_to_current_deployment_day() {
         .dashboard_observation
         .lock()
         .expect("dashboard observation") = Some(DashboardObservation {
+        runtime_slots: None,
         range: TimeRange::new(now - Duration::hours(1), now).expect("dashboard range"),
         totals: Default::default(),
         provider_accounts: AccountPoolMetrics::default(),
@@ -230,6 +233,22 @@ async fn usage_detail_should_keep_attempt_snapshot_contract() {
     let fixture = AdminTestFixture::new().await;
     fixture.auth.insert_session("valid-session");
     let now = Utc::now();
+    let mut record = usage_record_with_account(
+        "req_detail",
+        "acct_snap_a",
+        "Snapshot Alpha",
+        "alpha@example.invalid",
+        "oauth",
+        now,
+    );
+    record.first_token_ms = Some(20);
+    record.upstream_response_ms = Some(1_000);
+    record.upstream_api_overhead_ms = Some(120.25);
+    record.upstream_engine_ms = Some(6400.0);
+    record.upstream_engine_iapi_ttft_ms = Some(650.5);
+    record.upstream_engine_service_ttft_ms = Some(720.25);
+    record.upstream_engine_iapi_tbt_ms = Some(18.45);
+    record.upstream_engine_service_tbt_ms = Some(20.12);
     fixture
         .usage_detail
         .lock()
@@ -237,14 +256,7 @@ async fn usage_detail_should_keep_attempt_snapshot_contract() {
         .replace(UsageDetail {
             trace: None,
             related_requests: Vec::new(),
-            request: usage_record_with_account(
-                "req_detail",
-                "acct_snap_a",
-                "Snapshot Alpha",
-                "alpha@example.invalid",
-                "oauth",
-                now,
-            ),
+            request: record,
             attempts: vec![UsageAttempt {
                 source: "ops_event".to_owned(),
                 id: "ops_detail".to_owned(),
@@ -297,6 +309,39 @@ async fn usage_detail_should_keep_attempt_snapshot_contract() {
         .await
         .expect("usage detail body");
     let value: serde_json::Value = serde_json::from_slice(&body).expect("usage detail JSON");
+    assert_eq!(value["data"]["firstTokenLatencyMs"], 20);
+    assert_eq!(value["data"]["latencyDetails"]["upstreamResponseMs"], 1_000);
+    assert_eq!(
+        value["data"]["latencyDetails"]["upstreamApiOverheadMs"],
+        json!(120.25)
+    );
+    assert_eq!(
+        value["data"]["latencyDetails"]["upstreamEngineMs"],
+        json!(6400.0)
+    );
+    assert_eq!(
+        value["data"]["latencyDetails"]["upstreamEngineIapiTtftMs"],
+        json!(650.5)
+    );
+    assert_eq!(
+        value["data"]["latencyDetails"]["upstreamEngineServiceTtftMs"],
+        json!(720.25)
+    );
+    assert_eq!(
+        value["data"]["latencyDetails"]["upstreamEngineIapiTbtMs"],
+        json!(18.45)
+    );
+    assert_eq!(
+        value["data"]["latencyDetails"]["upstreamEngineServiceTbtMs"],
+        json!(20.12)
+    );
+    assert!(value["data"].get("firstTokenMs").is_none());
+    assert!(value["data"].get("firstTokenLatencyMsDisplay").is_none());
+    assert!(
+        value["data"]["latencyDetails"]
+            .get("firstTokenMs")
+            .is_none()
+    );
     assert_eq!(
         serde_json::json!({
             "accountId": value["data"]["accountId"],
@@ -699,6 +744,13 @@ fn usage_record_with_account(
         first_text_ms: None,
         first_token_ms: None,
         provider_processing_ms: None,
+        upstream_response_ms: None,
+        upstream_api_overhead_ms: None,
+        upstream_engine_ms: None,
+        upstream_engine_iapi_ttft_ms: None,
+        upstream_engine_service_ttft_ms: None,
+        upstream_engine_iapi_tbt_ms: None,
+        upstream_engine_service_tbt_ms: None,
         latency_ms: None,
         admission_decision_ms: None,
         account_selection_wait_ms: None,
@@ -831,6 +883,13 @@ async fn usage_route_should_expose_table_facts_without_detail_payload() {
             first_text_ms: Some(19),
             first_token_ms: Some(20),
             provider_processing_ms: Some(21),
+            upstream_response_ms: Some(2_000),
+            upstream_api_overhead_ms: None,
+            upstream_engine_ms: None,
+            upstream_engine_iapi_ttft_ms: None,
+            upstream_engine_service_ttft_ms: None,
+            upstream_engine_iapi_tbt_ms: None,
+            upstream_engine_service_tbt_ms: None,
             latency_ms: Some(31),
             admission_decision_ms: Some(1),
             account_selection_wait_ms: Some(2),
@@ -967,8 +1026,8 @@ async fn usage_route_should_expose_table_facts_without_detail_payload() {
                 "firstEventMs": 17,
                 "firstReasoningMs": 18,
                 "firstTextMs": 19,
-                "firstTokenMs": 20,
                 "openaiProcessingMs": 21,
+                "upstreamResponseMs": 2000,
             },
             "firstTokenLatencyMs": 20,
             "latencyMs": 31,
@@ -1042,6 +1101,7 @@ async fn calendar_queries_and_display_follow_deployment_timezone_with_one_anchor
         let fixture = AdminTestFixture::with_timezone(timezone).await;
         fixture.auth.insert_session("valid-session");
         *fixture.dashboard_observation.lock().unwrap() = Some(DashboardObservation {
+            runtime_slots: None,
             range,
             totals: Default::default(),
             provider_accounts: Default::default(),

backend/crates/gateway-api/tests/admin/plugins/management.rs

@@ -5,24 +5,31 @@ use axum::{
     http::{Request, StatusCode},
 };
 use futures::future::BoxFuture;
-use gateway_core::{
-    engine::{
-        execution::{
-            AuthenticatedClient, ClientAuthenticationError, ExecutionService,
-            PreparedRootExecution, StartExecution, StartProviderExecution, StartedExecution,
-        },
-        middleware::{
-            FrozenMiddlewarePlan, MiddlewareBody, MiddlewareContext, MiddlewareError,
-            MiddlewareFrame, MiddlewareFraming, MiddlewareHeader, MiddlewareNext, MiddlewarePlan,
-            MiddlewareRequest, MiddlewareResponse,
-        },
-    },
-    error::GatewayError,
-    lifecycle::CancellationToken,
-    policy::ClientApiKeyId,
-    routing::PublicModelId,
-    runtime::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference},
-};
+use gateway_core::engine::execution::AuthenticatedClient;
+use gateway_core::engine::execution::ClientAuthenticationError;
+use gateway_core::engine::execution::ExecutionService;
+use gateway_core::engine::execution::PreparedRootExecution;
+use gateway_core::engine::execution::StartExecution;
+use gateway_core::engine::execution::StartProviderExecution;
+use gateway_core::engine::execution::StartedExecution;
+use gateway_core::engine::middleware::FrozenMiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareBody;
+use gateway_core::engine::middleware::MiddlewareContext;
+use gateway_core::engine::middleware::MiddlewareError;
+use gateway_core::engine::middleware::MiddlewareFrame;
+use gateway_core::engine::middleware::MiddlewareFraming;
+use gateway_core::engine::middleware::MiddlewareHeader;
+use gateway_core::engine::middleware::MiddlewareNext;
+use gateway_core::engine::middleware::MiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareRequest;
+use gateway_core::engine::middleware::MiddlewareResponse;
+use gateway_core::error::GatewayError;
+use gateway_core::lifecycle::CancellationToken;
+use gateway_core::policy::ClientApiKeyId;
+use gateway_core::routing::PublicModelId;
+use gateway_core::routing::extensions::ExtensionSetId;
+use gateway_core::routing::extensions::ExtensionSetLease;
+use gateway_core::routing::extensions::ExtensionSetReference;
 use std::sync::{
     Arc, Mutex,
     atomic::{AtomicBool, Ordering},
@@ -368,10 +375,11 @@ pub(super) fn resource_fixture(
 }
 
 async fn published_fixture() -> AdminTestFixture {
-    use gateway_core::{
-        routing::{ConfigRevision, RuntimeSnapshot},
-        runtime::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference},
-    };
+    use gateway_core::routing::ConfigRevision;
+    use gateway_core::routing::RuntimeSnapshot;
+    use gateway_core::routing::extensions::ExtensionSetId;
+    use gateway_core::routing::extensions::ExtensionSetLease;
+    use gateway_core::routing::extensions::ExtensionSetReference;
     use std::sync::Arc;
     struct Lease;
     impl ExtensionSetLease for Lease {

backend/crates/gateway-api/tests/admin/plugins/mod.rs

@@ -546,7 +546,7 @@ pub(super) struct TestPluginPorts {
 impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPorts {
     async fn validate_target(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
     ) -> Result<(), AdminError> {
         Ok(())
@@ -554,7 +554,7 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
 
     async fn start_callback(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         _: gateway_admin::model::plugins::management::StartPluginManagementCallback,
         _: &gateway_admin::model::auth::AdminRequestContext,
@@ -564,7 +564,7 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
     }
     async fn callback(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         _: &str,
         request: gateway_admin::model::plugins::management::PluginManagementRequest,
@@ -576,14 +576,14 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
     }
     async fn views(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
     ) -> Result<Vec<gateway_admin::model::plugins::management::PluginManagementView>, AdminError>
     {
         Ok(Vec::new())
     }
     async fn resource(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         path: &str,
         public: bool,
@@ -596,7 +596,7 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
     }
     async fn handle(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::management::PluginManagementTarget,
         request: gateway_admin::model::plugins::management::PluginManagementRequest,
     ) -> Result<gateway_admin::model::plugins::management::PluginManagementResponse, AdminError>
@@ -611,7 +611,7 @@ impl gateway_admin::ports::plugin_management::PluginManagement for TestPluginPor
 impl gateway_admin::ports::plugins::PluginStateLifecycle for TestPluginPorts {
     async fn activate_state(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: &gateway_admin::model::plugins::instances::PluginInstance,
     ) -> Result<(), AdminError> {
         Ok(())
@@ -621,7 +621,7 @@ impl gateway_admin::ports::plugins::PluginStateLifecycle for TestPluginPorts {
     }
     async fn migrate_state(
         &self,
-        _: &gateway_core::runtime::extensions::ExtensionSetReference,
+        _: &gateway_core::routing::extensions::ExtensionSetReference,
         _: gateway_admin::model::plugins::state::PluginStateTransition,
     ) -> Result<(), AdminError> {
         panic!("unexpected state migration")
@@ -647,7 +647,7 @@ impl gateway_admin::ports::plugins::PluginPreparation for TestPluginPorts {
     async fn prepare(
         &self,
         _: gateway_admin::model::plugins::instances::PluginInstanceSnapshot,
-    ) -> Result<gateway_core::runtime::extensions::ExtensionSetReference, AdminError> {
+    ) -> Result<gateway_core::routing::extensions::ExtensionSetReference, AdminError> {
         Err(AdminError::invalid("unused plugin fixture"))
     }
 }
@@ -658,7 +658,7 @@ impl gateway_admin::ports::plugins::PluginRuntimeDiagnostics for TestPluginPorts
         &self,
         snapshot: &gateway_admin::model::plugins::instances::PluginInstanceSnapshot,
         _: Option<u64>,
-        _: Option<&gateway_core::runtime::extensions::ExtensionSetReference>,
+        _: Option<&gateway_core::routing::extensions::ExtensionSetReference>,
     ) -> Option<BTreeMap<String, gateway_admin::model::plugins::instances::PluginInstanceRuntime>>
     {
         Some(

backend/crates/gateway-api/tests/architecture.rs

@@ -1,4 +1,4 @@
-//! 检查 API 层依赖、源码与测试模块边界，以及管理路由约定
+//! 检查 API 层依赖与管理路由约定，模块布局和测试镜像由 workspace 架构检查覆盖
 
 use std::{fs, path::Path};
 
@@ -29,150 +29,6 @@ fn workspace_should_include_gateway_api() {
     assert!(workspace.contains("\"crates/gateway-api\""));
 }
 
-#[test]
-fn source_tree_should_match_frozen_machine_manifest() {
-    let root = Path::new(env!("CARGO_MANIFEST_DIR"));
-    let mut actual = rust_files(&root.join("src"));
-    actual.sort();
-    let mut expected = vec![
-        "src/admin/account_groups.rs",
-        "src/admin/accounts/credentials.rs",
-        "src/admin/accounts/handlers.rs",
-        "src/admin/accounts/import_tasks.rs",
-        "src/admin/accounts/mod.rs",
-        "src/admin/accounts/presenter.rs",
-        "src/admin/accounts/wire.rs",
-        "src/admin/auth.rs",
-        "src/admin/backups.rs",
-        "src/admin/client_keys.rs",
-        "src/admin/diagnostics.rs",
-        "src/admin/extract.rs",
-        "src/admin/mod.rs",
-        "src/admin/observability/mod.rs",
-        "src/admin/observability/presenter.rs",
-        "src/admin/observability/query.rs",
-        "src/admin/observability/routes.rs",
-        "src/admin/observability/wire.rs",
-        "src/admin/plugins/artifacts.rs",
-        "src/admin/plugins/distribution.rs",
-        "src/admin/plugins/instances.rs",
-        "src/admin/plugins/management.rs",
-        "src/admin/plugins/mod.rs",
-        "src/admin/presenter.rs",
-        "src/admin/proxies.rs",
-        "src/admin/settings.rs",
-        "src/admin/system.rs",
-        "src/admin/wire.rs",
-        "src/auth.rs",
-        "src/health.rs",
-        "src/key_usage/mod.rs",
-        "src/key_usage/query.rs",
-        "src/key_usage/presenter.rs",
-        "src/lib.rs",
-        "src/middleware/headers.rs",
-        "src/middleware/mod.rs",
-        "src/middleware/websocket.rs",
-        "src/openai/middleware/http.rs",
-        "src/openai/middleware/mod.rs",
-        "src/openai/middleware/request.rs",
-        "src/openai/middleware/session.rs",
-        "src/openai/auth.rs",
-        "src/openai/error.rs",
-        "src/openai/images/http.rs",
-        "src/openai/images/mod.rs",
-        "src/openai/live/http.rs",
-        "src/openai/live/mod.rs",
-        "src/openai/live/websocket.rs",
-        "src/openai/mod.rs",
-        "src/openai/models.rs",
-        "src/openai/endpoint.rs",
-        "src/openai/responses/error.rs",
-        "src/openai/responses/http.rs",
-        "src/openai/responses/mod.rs",
-        "src/openai/responses/request.rs",
-        "src/openai/responses/response.rs",
-        "src/openai/responses/validation.rs",
-        "src/openai/responses/websocket/connection/mod.rs",
-        "src/openai/responses/websocket/connection/middleware.rs",
-        "src/openai/responses/websocket/forward.rs",
-        "src/openai/responses/websocket/mod.rs",
-        "src/openai/responses/websocket/protocol.rs",
-        "src/openai/router.rs",
-        "src/openai/search/http.rs",
-        "src/openai/search/mod.rs",
-        "src/openai/service.rs",
-        "src/openai/usage.rs",
-        "src/provider.rs",
-        "src/session_cookie.rs",
-        "src/time.rs",
-    ];
-    expected.sort_unstable();
-
-    assert_eq!(actual, expected);
-}
-
-#[test]
-fn test_tree_should_match_frozen_rust_mirror() {
-    let root = Path::new(env!("CARGO_MANIFEST_DIR"));
-    let mut actual = rust_files(&root.join("tests"));
-    actual.sort();
-    let mut expected = vec![
-        "tests/admin/account_groups.rs",
-        "tests/admin/accounts/handlers.rs",
-        "tests/admin/accounts/import_tasks.rs",
-        "tests/admin/accounts/mod.rs",
-        "tests/admin/accounts/presenter.rs",
-        "tests/admin/auth.rs",
-        "tests/admin/client_keys.rs",
-        "tests/admin/errors.rs",
-        "tests/admin/error_diagnostics.rs",
-        "tests/admin/mod.rs",
-        "tests/admin/observability/mod.rs",
-        "tests/admin/observability/query.rs",
-        "tests/admin/observability/response.rs",
-        "tests/admin/plugins/management.rs",
-        "tests/admin/plugins/mod.rs",
-        "tests/admin/proxies.rs",
-        "tests/admin/settings.rs",
-        "tests/admin/system.rs",
-        "tests/admin/wire.rs",
-        "tests/architecture.rs",
-        "tests/auth.rs",
-        "tests/support.rs",
-        "tests/health.rs",
-        "tests/key_usage/mod.rs",
-        "tests/key_usage/fixtures.rs",
-        "tests/main.rs",
-        "tests/middleware/mod.rs",
-        "tests/middleware/websocket.rs",
-        "tests/openai/middleware/mod.rs",
-        "tests/openai/auth.rs",
-        "tests/openai/error.rs",
-        "tests/openai/images/mod.rs",
-        "tests/openai/mod.rs",
-        "tests/openai/models.rs",
-        "tests/openai/endpoint.rs",
-        "tests/openai/responses/http.rs",
-        "tests/openai/responses/mod.rs",
-        "tests/openai/responses/request.rs",
-        "tests/openai/responses/websocket/connection/mod.rs",
-        "tests/openai/responses/websocket/connection/middleware.rs",
-        "tests/openai/responses/websocket/forward.rs",
-        "tests/openai/responses/websocket/interrupt.rs",
-        "tests/openai/responses/websocket/mod.rs",
-        "tests/openai/responses/websocket/protocol.rs",
-        "tests/openai/live/mod.rs",
-        "tests/openai/live/websocket.rs",
-        "tests/openai/router.rs",
-        "tests/openai/search/mod.rs",
-        "tests/openai/usage.rs",
-        "tests/provider.rs",
-    ];
-    expected.sort_unstable();
-
-    assert_eq!(actual, expected);
-}
-
 #[test]
 fn admin_routes_should_use_only_get_post_and_static_paths() {
     let root = Path::new(env!("CARGO_MANIFEST_DIR")).join("src/admin");
@@ -192,20 +48,6 @@ fn admin_routes_should_use_only_get_post_and_static_paths() {
     );
 }
 
-fn rust_files(root: &Path) -> Vec<String> {
-    let manifest = Path::new(env!("CARGO_MANIFEST_DIR"));
-    all_files(root)
-        .into_iter()
-        .filter(|path| path.extension().and_then(|value| value.to_str()) == Some("rs"))
-        .map(|path| {
-            path.strip_prefix(manifest)
-                .expect("path below manifest")
-                .to_string_lossy()
-                .replace('\\', "/")
-        })
-        .collect()
-}
-
 fn all_files(root: &Path) -> Vec<std::path::PathBuf> {
     let mut files = Vec::new();
     let mut pending = vec![root.to_path_buf()];

backend/crates/gateway-api/tests/key_usage/fixtures.rs

@@ -135,6 +135,13 @@ fn usage_record() -> UsageListRecord {
         first_text_ms: Some(600),
         first_token_ms: Some(210),
         provider_processing_ms: None,
+        upstream_response_ms: Some(1_000),
+        upstream_api_overhead_ms: None,
+        upstream_engine_ms: None,
+        upstream_engine_iapi_ttft_ms: None,
+        upstream_engine_service_ttft_ms: None,
+        upstream_engine_iapi_tbt_ms: Some(2.450638),
+        upstream_engine_service_tbt_ms: None,
         latency_ms: Some(1250),
         admission_decision_ms: None,
         account_selection_wait_ms: None,

backend/crates/gateway-api/tests/key_usage/mod.rs

@@ -305,9 +305,20 @@ async fn records_keep_pagination_and_hide_admin_and_upstream_data() {
             assert_eq!(record["firstTokenLatencyMs"], 210);
             assert_fields(
                 &record["latencyDetails"],
-                &["firstEventMs", "firstReasoningMs", "firstTextMs"],
+                &[
+                    "upstreamResponseMs",
+                    "upstreamEngineIapiTbtMs",
+                    "firstEventMs",
+                    "firstReasoningMs",
+                    "firstTextMs",
+                ],
             );
             assert_eq!(record["latencyDetails"]["firstEventMs"], 100);
+            assert_eq!(record["latencyDetails"]["upstreamResponseMs"], 1000);
+            assert_eq!(
+                record["latencyDetails"]["upstreamEngineIapiTbtMs"],
+                2.450638
+            );
             assert!(record["statusCode"].is_null());
         } else {
             assert_eq!(record["statusCode"], 502);

backend/crates/gateway-api/tests/middleware/mod.rs

@@ -16,15 +16,18 @@ use axum::{
 };
 use bytes::Bytes;
 use futures::{future::BoxFuture, stream};
-use gateway_core::{
-    engine::middleware::{
-        FrozenMiddlewarePlan, MiddlewareContext, MiddlewareError, MiddlewareNext, MiddlewarePlan,
-        MiddlewareRequest, MiddlewareResponse,
-    },
-    lifecycle::CancellationToken,
-    middleware::http as contract,
-    runtime::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference},
-};
+use gateway_core::engine::middleware::FrozenMiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareContext;
+use gateway_core::engine::middleware::MiddlewareError;
+use gateway_core::engine::middleware::MiddlewareNext;
+use gateway_core::engine::middleware::MiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareRequest;
+use gateway_core::engine::middleware::MiddlewareResponse;
+use gateway_core::engine::middleware::http as contract;
+use gateway_core::lifecycle::CancellationToken;
+use gateway_core::routing::extensions::ExtensionSetId;
+use gateway_core::routing::extensions::ExtensionSetLease;
+use gateway_core::routing::extensions::ExtensionSetReference;
 use http_body::Frame;
 use http_body_util::{BodyExt as _, StreamBody};
 use tower::ServiceExt as _;
@@ -70,12 +73,12 @@ impl MiddlewarePlan for Plan {
     }
     fn handle_websocket(
         &self,
-        context: gateway_core::middleware::websocket::Context,
-        message: gateway_core::middleware::websocket::Message,
-        next: gateway_core::middleware::websocket::Next,
+        context: gateway_core::engine::middleware::websocket::Context,
+        message: gateway_core::engine::middleware::websocket::Message,
+        next: gateway_core::engine::middleware::websocket::Next,
     ) -> BoxFuture<
         'static,
-        Result<Option<gateway_core::middleware::websocket::Message>, MiddlewareError>,
+        Result<Option<gateway_core::engine::middleware::websocket::Message>, MiddlewareError>,
     > {
         assert!(context.plan.is_some());
         self.calls.fetch_add(1, Ordering::SeqCst);
@@ -103,11 +106,13 @@ impl MiddlewarePlan for Plan {
     }
     fn handle_service(
         &self,
-        context: gateway_core::middleware::service::Context,
+        context: gateway_core::engine::middleware::service::Context,
         input: serde_json::Value,
-        next: gateway_core::middleware::service::Next,
-    ) -> BoxFuture<'static, Result<serde_json::Value, gateway_core::middleware::service::Error>>
-    {
+        next: gateway_core::engine::middleware::service::Next,
+    ) -> BoxFuture<
+        'static,
+        Result<serde_json::Value, gateway_core::engine::middleware::service::Error>,
+    > {
         self.calls.fetch_add(1, Ordering::SeqCst);
         assert_eq!(context.operation, "settings.load");
         assert_eq!(context.request_id, "service-request");
@@ -146,7 +151,7 @@ impl MiddlewarePlan for Plan {
                     response.extensions_mut().insert(contract::upgrade::Upgraded::new(status, Box::pin(async move {
                         let session = session.await.map_err(|_| MiddlewareError::Fault)?;
                         if idle { return std::future::pending().await; }
-                        use gateway_core::middleware::websocket::{Kind, Message};
+                        use gateway_core::engine::middleware::websocket::{Kind, Message};
                         let receive = session.receive();
                         tokio::pin!(receive);
                         tokio::select! {

backend/crates/gateway-api/tests/openai/auth.rs

@@ -11,9 +11,7 @@ use axum::http::{
 };
 use futures::future::BoxFuture;
 
-use gateway_api::openai::auth::{
-    ClientApiKeyAuthError, bearer_client_api_key, identify_codex_client,
-};
+use gateway_api::openai::auth::{ClientApiKeyAuthError, identify_codex_client};
 use gateway_core::{
     engine::{
         authentication::ClientAuthenticationRequest,
@@ -118,114 +116,76 @@ async fn plugin_authentication_envelope_contains_only_authorization() {
     assert_eq!(execution.calls.load(Ordering::SeqCst), 1);
 }
 
-#[test]
-fn bearer_client_api_key_should_reject_missing_authorization() {
-    assert_eq!(
-        bearer_client_api_key(&HeaderMap::new()),
-        Err(ClientApiKeyAuthError::MissingAuthorization)
+#[tokio::test]
+async fn native_authentication_checks_the_current_envelope_at_the_http_entry() {
+    let fixture = crate::admin::AdminTestFixture::new().await;
+    let maximum_key = "x".repeat(
+        gateway_core::engine::authentication::MAXIMUM_AUTHORIZATION_BYTES - "Bearer ".len(),
     );
-}
-
-#[test]
-fn actor_authorization_marker_should_not_replace_client_api_key() {
-    let mut headers = HeaderMap::new();
-    headers.insert(
-        "x-openai-actor-authorization",
-        HeaderValue::from_static("proxy-managed"),
-    );
-
-    assert_eq!(
-        bearer_client_api_key(&headers),
-        Err(ClientApiKeyAuthError::MissingAuthorization)
-    );
-}
-
-#[test]
-fn actor_authorization_marker_should_preserve_existing_bearer_authentication() {
-    let mut headers = HeaderMap::new();
-    headers.insert(
-        "x-openai-actor-authorization",
-        HeaderValue::from_static("proxy-managed"),
-    );
-    headers.insert(AUTHORIZATION, HeaderValue::from_static("Bearer sk_client"));
-
-    assert_eq!(bearer_client_api_key(&headers), Ok("sk_client"));
-}
-
-#[test]
-fn bearer_client_api_key_should_reject_non_utf8_authorization() {
-    let mut headers = HeaderMap::new();
-    headers.insert(
-        AUTHORIZATION,
-        HeaderValue::from_bytes(&[0xff]).expect("opaque header value"),
-    );
-
-    assert_eq!(
-        bearer_client_api_key(&headers),
-        Err(ClientApiKeyAuthError::MalformedAuthorization)
-    );
-}
-
-#[test]
-fn bearer_client_api_key_should_require_exact_bearer_scheme() {
-    let mut headers = HeaderMap::new();
-    headers.insert(AUTHORIZATION, HeaderValue::from_static("bearer sk_client"));
-
-    assert_eq!(
-        bearer_client_api_key(&headers),
-        Err(ClientApiKeyAuthError::MalformedAuthorization)
-    );
-}
-
-#[test]
-fn bearer_client_api_key_should_reject_empty_bearer_token() {
-    let mut headers = HeaderMap::new();
-    headers.insert(AUTHORIZATION, HeaderValue::from_static("Bearer    "));
-
-    assert_eq!(
-        bearer_client_api_key(&headers),
-        Err(ClientApiKeyAuthError::MalformedAuthorization)
-    );
-}
-
-#[test]
-fn bearer_client_api_key_should_accept_migrated_formats_without_rewriting() {
-    for key in [
-        "q".to_owned(),
-        "sk-old/key+value=:!@".to_owned(),
-        "x".repeat(8192),
-    ] {
-        let mut headers = HeaderMap::new();
-        headers.insert(
-            AUTHORIZATION,
-            HeaderValue::from_str(&format!("Bearer {key}")).unwrap(),
-        );
-        assert_eq!(bearer_client_api_key(&headers), Ok(key.as_str()));
+    let oversized_key = format!("{maximum_key}x");
+    let mut cases = vec![
+        ("missing", "sk_client", None, StatusCode::UNAUTHORIZED),
+        (
+            "opaque",
+            "sk_client",
+            Some(HeaderValue::from_bytes(&[0xff]).unwrap()),
+            StatusCode::UNAUTHORIZED,
+        ),
+        (
+            "lowercase scheme",
+            "sk_client",
+            Some(HeaderValue::from_static("bearer sk_client")),
+            StatusCode::UNAUTHORIZED,
+        ),
+        (
+            "empty token",
+            "sk_client",
+            Some(HeaderValue::from_static("Bearer    ")),
+            StatusCode::UNAUTHORIZED,
+        ),
+        (
+            "whitespace inside token",
+            "sk_client",
+            Some(HeaderValue::from_static("Bearer key with spaces")),
+            StatusCode::UNAUTHORIZED,
+        ),
+        (
+            "trimmed token",
+            "sk_client",
+            Some(HeaderValue::from_static("Bearer   sk_client   ")),
+            StatusCode::OK,
+        ),
+    ];
+    for key in ["q", "sk-old/key+value=:!@", maximum_key.as_str()] {
+        cases.push((
+            "configured native key",
+            key,
+            Some(HeaderValue::from_str(&format!("Bearer {key}")).unwrap()),
+            StatusCode::OK,
+        ));
     }
-}
+    cases.push((
+        "oversized envelope",
+        oversized_key.as_str(),
+        Some(HeaderValue::from_str(&format!("Bearer {oversized_key}")).unwrap()),
+        StatusCode::UNAUTHORIZED,
+    ));
 
-#[test]
-fn bearer_client_api_key_should_reject_whitespace_inside_tokens() {
-    let mut headers = HeaderMap::new();
-    headers.insert(
-        AUTHORIZATION,
-        HeaderValue::from_static("Bearer key with spaces"),
-    );
-    assert_eq!(
-        bearer_client_api_key(&headers),
-        Err(ClientApiKeyAuthError::InvalidKeyFormat)
-    );
-}
-
-#[test]
-fn bearer_client_api_key_should_return_trimmed_gateway_key() {
-    let mut headers = HeaderMap::new();
-    headers.insert(
-        AUTHORIZATION,
-        HeaderValue::from_static("Bearer   sk_client_secret   "),
-    );
-
-    assert_eq!(bearer_client_api_key(&headers), Ok("sk_client_secret"));
+    for (case, key, authorization, expected) in cases {
+        let app =
+            super::api_router_with_admin_and_client(fixture.services.clone(), key, "key-auth");
+        let mut request = crate::support::empty_request(Method::GET, "/v1/models");
+        // actor 标记既不能代替认证，也不能干扰已有的原生 Key
+        request.headers_mut().insert(
+            "x-openai-actor-authorization",
+            HeaderValue::from_static("proxy-managed"),
+        );
+        if let Some(authorization) = authorization {
+            request.headers_mut().insert(AUTHORIZATION, authorization);
+        }
+        let response = app.oneshot(request).await.unwrap();
+        assert_eq!(response.status(), expected, "{case}");
+    }
 }
 
 #[test]
@@ -414,9 +374,9 @@ fn chatgpt_remote_suffix_created_by_header_truncation_should_not_skip_the_gate()
 
 #[tokio::test]
 async fn http_settings_freeze_before_plan_resolution_and_apply_before_admission() {
-    use gateway_core::{
-        engine::middleware::*, middleware::http as contract, runtime::extensions::*,
-    };
+    use gateway_core::engine::middleware::http as contract;
+    use gateway_core::engine::middleware::*;
+    use gateway_core::routing::extensions::*;
     struct Lease;
     impl ExtensionSetLease for Lease {
         fn is_ready(&self) -> bool {

backend/crates/gateway-api/tests/openai/middleware/mod.rs

@@ -10,15 +10,22 @@ use std::{
 
 use bytes::Bytes;
 use futures::future::BoxFuture;
-use gateway_core::{
-    engine::middleware::{
-        FrozenMiddlewarePlan, MiddlewareBody, MiddlewareContext, MiddlewareError, MiddlewareFrame,
-        MiddlewareFraming, MiddlewareHeader, MiddlewareMount, MiddlewareNext, MiddlewarePlan,
-        MiddlewareRequest, MiddlewareResponse,
-    },
-    operation::OperationKind,
-    runtime::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference},
-};
+use gateway_core::engine::middleware::FrozenMiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareBody;
+use gateway_core::engine::middleware::MiddlewareContext;
+use gateway_core::engine::middleware::MiddlewareError;
+use gateway_core::engine::middleware::MiddlewareFrame;
+use gateway_core::engine::middleware::MiddlewareFraming;
+use gateway_core::engine::middleware::MiddlewareHeader;
+use gateway_core::engine::middleware::MiddlewareMount;
+use gateway_core::engine::middleware::MiddlewareNext;
+use gateway_core::engine::middleware::MiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareRequest;
+use gateway_core::engine::middleware::MiddlewareResponse;
+use gateway_core::operation::OperationKind;
+use gateway_core::routing::extensions::ExtensionSetId;
+use gateway_core::routing::extensions::ExtensionSetLease;
+use gateway_core::routing::extensions::ExtensionSetReference;
 
 /// OpenAI API 入口共用的中间件夹具；授权和 RPC 由 Runtime 行为测试覆盖
 #[derive(Debug, Default)]
@@ -30,7 +37,8 @@ pub(super) struct RequestMiddleware {
     pub(super) response_actions: Mutex<VecDeque<ResponseFrameAction>>,
     pub(super) response_headers: Vec<MiddlewareHeader>,
     pub(super) short_circuit_json: Option<Bytes>,
-    pub(super) settings: Option<fn(&mut gateway_core::settings::ExecutionSettings)>,
+    pub(super) settings:
+        Option<fn(&mut gateway_core::routing::request_settings::ExecutionSettings)>,
 }
 
 #[derive(Debug, Clone)]

backend/crates/gateway-api/tests/openai/responses/http.rs

@@ -2279,6 +2279,7 @@ async fn buffered_response_commits_only_after_complete_json_is_encoded() {
 async fn buffered_response_collects_completed_output_items_without_rewriting_streams() {
     let message = json!({
         "id": "msg_test", "type": "message", "role": "assistant", "status": "completed",
+        "phase": "partial_answer",
         "content": [{"type": "output_text", "text": "完整文本", "annotations": []}],
         "future_item_field": {"keep": true}
     });
@@ -2288,6 +2289,7 @@ async fn buffered_response_collects_completed_output_items_without_rewriting_str
     });
     let terminal = json!({
         "id": "resp_test", "status": "completed", "output": [],
+        "end_turn": false,
         "usage": {"input_tokens": 20, "output_tokens": 8, "total_tokens": 28},
         "future_terminal_field": {"keep": true}
     });

backend/crates/gateway-api/tests/openai/responses/websocket/connection/middleware.rs

@@ -2,14 +2,100 @@
 
 use super::*;
 use bytes::Bytes;
-use gateway_core::{
-    engine::middleware::{
-        FrozenMiddlewarePlan, MiddlewareContext, MiddlewareError, MiddlewareNext, MiddlewarePlan,
-        MiddlewareRequest, MiddlewareResponse,
-    },
-    middleware::{http as http_contract, websocket as core},
-    runtime::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference},
-};
+use gateway_core::engine::middleware::FrozenMiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareContext;
+use gateway_core::engine::middleware::MiddlewareError;
+use gateway_core::engine::middleware::MiddlewareNext;
+use gateway_core::engine::middleware::MiddlewarePlan;
+use gateway_core::engine::middleware::MiddlewareRequest;
+use gateway_core::engine::middleware::MiddlewareResponse;
+use gateway_core::engine::middleware::http as http_contract;
+use gateway_core::engine::middleware::websocket as core;
+use gateway_core::routing::extensions::ExtensionSetId;
+use gateway_core::routing::extensions::ExtensionSetLease;
+use gateway_core::routing::extensions::ExtensionSetReference;
+
+#[tokio::test(start_paused = true)]
+async fn client_close_does_not_flush_a_retained_sender_with_an_inflight_write() {
+    #[derive(Default)]
+    struct RetainedSenderPlan(Mutex<Option<Arc<dyn core::Sender>>>);
+    impl fmt::Debug for RetainedSenderPlan {
+        fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+            formatter.write_str("RetainedSenderPlan")
+        }
+    }
+    impl MiddlewarePlan for RetainedSenderPlan {
+        fn has_websocket(&self) -> bool {
+            true
+        }
+        fn handle_websocket(
+            &self,
+            context: core::Context,
+            message: core::Message,
+            next: core::Next,
+        ) -> BoxFuture<'static, Result<Option<core::Message>, MiddlewareError>> {
+            *self.0.lock().unwrap() = Some(context.sender);
+            next.run(message)
+        }
+        fn handle(
+            &self,
+            _: MiddlewareContext,
+            request: MiddlewareRequest,
+            next: MiddlewareNext,
+        ) -> BoxFuture<'static, Result<MiddlewareResponse, MiddlewareError>> {
+            next.run(request)
+        }
+    }
+
+    let plan = Arc::new(RetainedSenderPlan::default());
+    let frozen = FrozenMiddlewarePlan::new(
+        plan.clone(),
+        ExtensionSetReference::new(
+            ExtensionSetId::new("retained-writer".into()).unwrap(),
+            Arc::new(Lease),
+        ),
+    );
+    let (incoming, received) = unbounded_channel();
+    let (written, mut output) = unbounded_channel();
+    let flush_started = Arc::new(AtomicBool::new(false));
+    let socket = TestSocket {
+        incoming: received,
+        written,
+        stall_writes: true,
+        flush_started: flush_started.clone(),
+        dropped: Arc::new(AtomicBool::new(false)),
+    };
+    let mut connection = spawn_connection(
+        socket,
+        Arc::from("retained-writer"),
+        CancellationToken::new(),
+        ConnectionConfig::PRODUCTION,
+        Some(frozen),
+        Arc::from([]),
+    );
+    incoming.send(Ok(Message::Text("capture".into()))).unwrap();
+    assert!(matches!(
+        connection.next_event().await,
+        Some(ConnectionEvent::Text(_))
+    ));
+    let sender = plan.0.lock().unwrap().clone().unwrap();
+    let mut write = sender.send(core::Message {
+        kind: core::Kind::Text,
+        payload: Bytes::from_static(b"pending"),
+    });
+    assert!(futures::poll!(write.as_mut()).is_pending());
+    // 写 future 在 pump 的 select 之外持锁，而且暂不继续 poll 以响应取消
+    incoming.send(Ok(Message::Close(None))).unwrap();
+    let reason = tokio::time::timeout(Duration::from_millis(50), connection.wait_for_exit())
+        .await
+        .expect("peer close must not wait for a retained sender");
+    assert_eq!(reason, PumpExitReason::ClientClose);
+    assert!(!flush_started.load(Ordering::Acquire));
+    assert!(write.await.is_err());
+    assert!(matches!(output.try_recv(), Err(TryRecvError::Empty)));
+    plan.0.lock().unwrap().take();
+}
+
 #[derive(Debug, Default)]
 struct Plan {
     block_terminal: bool,

backend/crates/gateway-api/tests/openai/responses/websocket/connection/mod.rs

@@ -234,6 +234,7 @@ struct TestSocket {
     incoming: UnboundedReceiver<Result<Message, TestSocketError>>,
     written: UnboundedSender<Message>,
     stall_writes: bool,
+    flush_started: Arc<AtomicBool>,
     dropped: Arc<AtomicBool>,
 }
 
@@ -268,7 +269,12 @@ impl Sink<Message> for TestSocket {
     }
 
     fn poll_flush(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
-        Poll::Ready(Ok(()))
+        self.flush_started.store(true, Ordering::Release);
+        if self.stall_writes {
+            Poll::Pending
+        } else {
+            Poll::Ready(Ok(()))
+        }
     }
 
     fn poll_close(self: Pin<&mut Self>, _: &mut Context<'_>) -> Poll<Result<(), Self::Error>> {
@@ -286,6 +292,7 @@ struct PumpHarness {
     connection: ResponsesWebSocketConnection,
     incoming: UnboundedSender<Result<Message, TestSocketError>>,
     written: UnboundedReceiver<Message>,
+    flush_started: Arc<AtomicBool>,
     dropped: Arc<AtomicBool>,
     cancellation: CancellationToken,
 }
@@ -294,11 +301,13 @@ fn test_connection(stall_writes: bool) -> PumpHarness {
     let (incoming_tx, incoming_rx) = unbounded_channel();
     let (written_tx, written_rx) = unbounded_channel();
     let dropped = Arc::new(AtomicBool::new(false));
+    let flush_started = Arc::new(AtomicBool::new(false));
     let cancellation = CancellationToken::new();
     let socket = TestSocket {
         incoming: incoming_rx,
         written: written_tx,
         stall_writes,
+        flush_started: Arc::clone(&flush_started),
         dropped: Arc::clone(&dropped),
     };
     let connection = spawn_connection(
@@ -306,11 +315,14 @@ fn test_connection(stall_writes: bool) -> PumpHarness {
         Arc::from("ws_test"),
         cancellation.clone(),
         ConnectionConfig::PRODUCTION,
+        None,
+        Arc::from([]),
     );
     PumpHarness {
         connection,
         incoming: incoming_tx,
         written: written_rx,
+        flush_started,
         dropped,
         cancellation,
     }
@@ -402,6 +414,62 @@ async fn client_close_should_take_priority_over_queued_business_frames() {
     ));
 }
 
+#[tokio::test(start_paused = true)]
+async fn client_close_flush_has_a_short_deadline_without_sending_another_frame() {
+    let PumpHarness {
+        mut connection,
+        incoming,
+        mut written,
+        flush_started,
+        dropped,
+        ..
+    } = test_connection(true);
+    incoming.send(Ok(Message::Close(None))).expect("close peer");
+    let started = tokio::time::Instant::now();
+
+    let reason = tokio::time::timeout(Duration::from_secs(2), connection.wait_for_exit())
+        .await
+        .expect("peer close cannot consume the ordinary write timeout");
+    assert_eq!(reason, PumpExitReason::ClientClose);
+    assert_eq!(started.elapsed(), Duration::from_secs(1));
+    assert!(flush_started.load(Ordering::Acquire));
+    assert!(dropped.load(Ordering::Acquire));
+    assert!(
+        written.recv().await.is_none(),
+        "do not send a duplicate Close"
+    );
+}
+
+#[tokio::test(start_paused = true)]
+async fn lifecycle_shutdown_interrupts_a_pending_client_close_flush() {
+    let PumpHarness {
+        mut connection,
+        incoming,
+        flush_started,
+        cancellation,
+        dropped,
+        ..
+    } = test_connection(true);
+    incoming.send(Ok(Message::Close(None))).expect("close peer");
+    for _ in 0..100 {
+        if flush_started.load(Ordering::Acquire) {
+            break;
+        }
+        tokio::task::yield_now().await;
+    }
+    assert!(
+        flush_started.load(Ordering::Acquire),
+        "close flush must start"
+    );
+    cancellation.cancel();
+
+    let reason = tokio::time::timeout(Duration::from_millis(50), connection.wait_for_exit())
+        .await
+        .expect("host cancellation must not wait for the close timeout");
+    assert_eq!(reason, PumpExitReason::ClientClose);
+    assert!(dropped.load(Ordering::Acquire));
+}
+
 #[tokio::test]
 async fn full_business_queue_should_not_block_lifecycle_shutdown() {
     let PumpHarness {
@@ -986,6 +1054,40 @@ async fn requests_deferred_during_active_response_share_the_inbound_queue_limit(
     server.abort();
 }
 
+#[tokio::test]
+async fn client_normal_close_after_completed_response_receives_normal_close_reply() {
+    let (trace, mut socket, server) = start_active_response().await;
+    trace.release_terminal.notify_one();
+    loop {
+        let message = tokio::time::timeout(Duration::from_secs(2), socket.next())
+            .await
+            .expect("terminal response timeout")
+            .expect("response remains open")
+            .expect("valid response frame");
+        if let ClientMessage::Text(text) = message {
+            let event: Value = serde_json::from_str(&text).expect("response JSON");
+            if event["type"] == "response.completed" {
+                break;
+            }
+        }
+    }
+    socket
+        .close(Some(tokio_tungstenite::tungstenite::protocol::CloseFrame {
+            code: CloseCode::Normal,
+            reason: "client finished".into(),
+        }))
+        .await
+        .expect("send normal close");
+    // close() 只发送本端关闭帧；必须读取对端应答，才能证明握手正常完成
+    let reply = tokio::time::timeout(Duration::from_secs(2), socket.next()).await;
+    server.abort();
+    assert!(
+        matches!(&reply, Ok(Some(Ok(ClientMessage::Close(Some(frame)))))
+            if frame.code == CloseCode::Normal && frame.reason == "client finished"),
+        "client must receive the close acknowledgement, got {reply:?}"
+    );
+}
+
 #[tokio::test]
 async fn client_close_should_cancel_active_execution_without_starting_queued_requests() {
     let (trace, mut socket, server) = start_active_response().await;
@@ -1007,6 +1109,12 @@ async fn client_close_should_cancel_active_execution_without_starting_queued_req
     .expect("cancel active execution promptly");
     assert_eq!(trace.starts.load(Ordering::Acquire), 1);
     assert!(trace.cancelled.load(Ordering::Acquire));
+    let reply = tokio::time::timeout(Duration::from_secs(2), socket.next())
+        .await
+        .expect("close reply timeout")
+        .expect("close reply")
+        .expect("clean close handshake");
+    assert!(matches!(reply, ClientMessage::Close(None)));
     server.abort();
 }
 
@@ -1071,6 +1179,7 @@ async fn stalled_heartbeat_can_be_cancelled_without_waiting_for_write_timeout()
         written: _written,
         dropped,
         cancellation,
+        ..
     } = test_connection(true);
     tokio::time::advance(Duration::from_secs(25)).await;
     tokio::task::yield_now().await;

backend/crates/gateway-api/tests/openai/responses/websocket/forward.rs

@@ -19,7 +19,7 @@ use gateway_core::error::{
     ClientVisibleUpstreamError, ClientVisibleUpstreamResponse, ProviderError, ProviderErrorKind,
 };
 use gateway_core::event::{ProtocolWireEvent, ProviderEvent, ProviderResponseHeader};
-use gateway_core::runtime::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference};
+use gateway_core::routing::extensions::{ExtensionSetId, ExtensionSetLease, ExtensionSetReference};
 use gateway_core::upstream::{OpaqueUpstreamValue, UpstreamSendState};
 use serde_json::{Value, json};
 use tokio::net::TcpStream;

backend/crates/gateway-core/Cargo.toml

@@ -20,7 +20,6 @@ chrono-tz = { version = "0.10", features = ["serde"] }
 event-listener = "5.4.2"
 futures = "0.3.32"
 http = "1.5.0"
-http-body = "1.0.1"
 http-body-util = "0.1.5"
 futures-timer = "3.0.3"
 serde = { version = "1.0.228", features = ["derive", "rc"] }

backend/crates/gateway-core/src/account/model.rs

@@ -90,18 +90,6 @@ impl AccountConcurrency {
             Self::Limited(limit) => Some(limit),
         }
     }
-
-    /// 扣除当前请求不可占用的预留名额；不限并发不受影响，有限上限至少保留 1 个名额
-    #[must_use]
-    pub const fn excluding_reserved(self, reserved: u32) -> Self {
-        match self {
-            Self::Unlimited => Self::Unlimited,
-            Self::Limited(limit) => match NonZeroU32::new(limit.get().saturating_sub(reserved)) {
-                Some(remaining) => Self::Limited(remaining),
-                None => Self::Limited(NonZeroU32::MIN),
-            },
-        }
-    }
 }
 
 impl From<NonZeroU32> for AccountConcurrency {
@@ -588,18 +576,6 @@ impl AccountStatus {
             _ => None,
         }
     }
-
-    /// 管理页稳定业务排序：正常、限流、耗尽、错误、停用
-    #[must_use]
-    pub const fn sort_rank(self) -> u8 {
-        match self {
-            Self::Normal => 0,
-            Self::RateLimited => 1,
-            Self::QuotaExhausted => 2,
-            Self::Error => 3,
-            Self::Disabled => 4,
-        }
-    }
 }
 
 /// `error` 状态下的稳定原因码

backend/crates/gateway-core/src/account/selection.rs

@@ -206,6 +206,7 @@ impl AccountSelectionPolicy {
 /// Store 提供并发事实，Provider 叠加自己解释的额度事实；全部信号均可重建
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct AccountRuntimeSignals {
+    /// 本次调度容量池的在途数，不合并其他独立池
     pub in_flight: u32,
     pub last_started_at: Option<SystemTime>,
     pub quota_reset_at: Option<SystemTime>,
@@ -523,7 +524,7 @@ pub struct AccountSelectionContext {
     pub round_robin_cursor: u64,
     pub eligibility: AccountEligibilityPolicy,
     pub account_scope: Option<std::sync::Arc<crate::account::scope::FrozenAccountScope>>,
-    /// 本请求不可占用的每账号预留并发名额，由 Provider 按请求类别决定；0 表示不预留
+    /// 本请求使用的独立预留池上限，由 Provider 按请求类别决定；0 表示使用普通账号池
     pub reserved_concurrency: u32,
 }
 
@@ -531,9 +532,11 @@ impl AccountSelectionContext {
     /// 本请求在该账号上可使用的并发上限；资格判断、租约与策略投影必须共用这一口径
     #[must_use]
     pub fn concurrency_limit(&self, account: &ProviderAccount) -> AccountConcurrency {
-        account
-            .effective_concurrency(self.policy.max_concurrent_per_account())
-            .excluding_reserved(self.reserved_concurrency)
+        if self.reserved_concurrency > 0 {
+            AccountConcurrency::new(self.reserved_concurrency)
+        } else {
+            account.effective_concurrency(self.policy.max_concurrent_per_account())
+        }
     }
 }
 
@@ -723,7 +726,6 @@ impl AccountSelector {
 
         let candidate = match context.policy.strategy() {
             RotationStrategy::QuotaResetPriority => {
-                let default_concurrency = context.policy.max_concurrent_per_account();
                 eligible.sort_by(|left, right| {
                     (
                         left.signals.quota_reset_at.is_none(),
@@ -734,8 +736,8 @@ impl AccountSelector {
                             right.signals.quota_reset_at,
                         ))
                         .then_with(|| {
-                            capacity_utilization(left, default_concurrency)
-                                .total_cmp(&capacity_utilization(right, default_concurrency))
+                            capacity_utilization(left, context)
+                                .total_cmp(&capacity_utilization(right, context))
                         })
                         .then_with(|| {
                             left.signals
@@ -751,13 +753,7 @@ impl AccountSelector {
                 let index = context.round_robin_cursor as usize % eligible.len();
                 eligible.get(index).copied()?
             }
-            RotationStrategy::Smart => select_smart_candidate(
-                &eligible,
-                context.policy.max_concurrent_per_account(),
-                context.round_robin_cursor,
-                context.policy.smart_scheduling(),
-                context.now,
-            )?,
+            RotationStrategy::Smart => select_smart_candidate(&eligible, context)?,
             RotationStrategy::Sticky => {
                 eligible.sort_by_key(|candidate| {
                     (
@@ -843,17 +839,7 @@ impl AccountSelector {
         // 账号信号在队列锁外评分，锁内只读取实时队长并查表，避免扫描候选阻塞其他等待者
         let scores = candidates
             .iter()
-            .map(|candidate| {
-                (
-                    candidate.account.id(),
-                    smart_score(
-                        candidate,
-                        context.policy.max_concurrent_per_account(),
-                        config,
-                        context.now,
-                    ),
-                )
-            })
+            .map(|candidate| (candidate.account.id(), smart_score(candidate, context)))
             .collect::<HashMap<_, _>>();
         waiting
             .wait_with_priority(keys, |key, count| {
@@ -914,13 +900,9 @@ const SMART_LATENCY_HALF_SCORE_MS: f64 = 10_000.0;
 // 距重置一小时时得分减半；未知和已过期时间不提供重置奖励
 const SMART_RESET_HALF_SCORE_SECONDS: f64 = 3_600.0;
 
-fn capacity_utilization(
-    candidate: &AccountCandidate,
-    default_concurrency: AccountConcurrency,
-) -> f64 {
-    candidate
-        .account
-        .effective_concurrency(default_concurrency)
+fn capacity_utilization(candidate: &AccountCandidate, context: &AccountSelectionContext) -> f64 {
+    context
+        .concurrency_limit(&candidate.account)
         .limit()
         .map_or(0.0, |limit| {
             f64::from(candidate.signals.in_flight) / f64::from(limit.get())
@@ -929,38 +911,29 @@ fn capacity_utilization(
 
 fn select_smart_candidate<'a>(
     candidates: &[&'a AccountCandidate],
-    default_concurrency: AccountConcurrency,
-    cursor: u64,
-    config: SmartSchedulingConfig,
-    now: SystemTime,
+    context: &AccountSelectionContext,
 ) -> Option<&'a AccountCandidate> {
     let mut ranked = candidates
         .iter()
-        .map(|candidate| {
-            (
-                *candidate,
-                smart_score(candidate, default_concurrency, config, now),
-            )
-        })
+        .map(|candidate| (*candidate, smart_score(candidate, context)))
         .collect::<Vec<_>>();
     let best_score = ranked
         .iter()
         .map(|(_, score)| *score)
         .max_by(f64::total_cmp)?;
-    ranked.retain(|(_, score)| best_score - score <= config.score_tolerance());
+    ranked.retain(|(_, score)| {
+        best_score - score <= context.policy.smart_scheduling().score_tolerance()
+    });
     // 轮换顺序保持稳定，避免分数轻微交错与 cursor 同步后仍反复命中同一账号
     ranked.sort_unstable_by(|(left, _), (right, _)| left.account.id().cmp(right.account.id()));
-    let index = (cursor % ranked.len() as u64) as usize;
+    let index = (context.round_robin_cursor % ranked.len() as u64) as usize;
     Some(ranked[index].0)
 }
 
-pub(crate) fn smart_score(
-    candidate: &AccountCandidate,
-    default_concurrency: AccountConcurrency,
-    config: SmartSchedulingConfig,
-    now: SystemTime,
-) -> f64 {
-    let load = 1.0 - capacity_utilization(candidate, default_concurrency).clamp(0.0, 1.0);
+pub(crate) fn smart_score(candidate: &AccountCandidate, context: &AccountSelectionContext) -> f64 {
+    let config = context.policy.smart_scheduling();
+    let now = context.now;
+    let load = 1.0 - capacity_utilization(candidate, context).clamp(0.0, 1.0);
     let quota = candidate
         .signals
         .quota_remaining_rank

backend/crates/gateway-core/src/diagnostics/selection.rs

@@ -41,17 +41,14 @@ impl TraceContext {
                     "blocker": AccountSelector.scheduling_blocker(candidate, context)
                         .map(|blocker| format!("{blocker:?}")),
                     "inFlight": signals.in_flight,
-                    "concurrencyLimit": candidate.account
-                        .effective_concurrency(context.policy.max_concurrent_per_account()).get(),
+                    "concurrencyLimit": context.concurrency_limit(&candidate.account).get(),
                     "quotaRemainingBasisPoints": signals.quota_remaining_rank,
                     "quotaResetAtUnixMs": signals.quota_reset_at.and_then(|reset|
                         reset.duration_since(std::time::UNIX_EPOCH).ok()
                     ).map(|duration| duration.as_millis()),
                     "failureRateBasisPoints": signals.failure_rate_basis_points,
                     "firstOutputLatencyMs": signals.first_output_latency_ms,
-                    "smartScore": smart.then(|| smart_score(
-                        candidate, context.policy.max_concurrent_per_account(), context.policy.smart_scheduling(), context.now
-                    )),
+                    "smartScore": smart.then(|| smart_score(candidate, context)),
                 })
             })
             .collect::<Vec<_>>();

backend/crates/gateway-core/src/engine/authentication.rs

@@ -10,7 +10,7 @@ use futures::future::BoxFuture;
 
 use crate::{
     policy::{ClientApiKeyId, PlaintextClientApiKey},
-    runtime::extensions::{ExtensionSetId, ExtensionSetReference},
+    routing::extensions::{ExtensionSetId, ExtensionSetReference},
 };
 
 pub const MAXIMUM_AUTHORIZATION_BYTES: usize = 8 * 1024;
@@ -18,24 +18,29 @@ pub const MAXIMUM_AUTHORIZATION_BYTES: usize = 8 * 1024;
 /// API 从数据面 Authorization 头构造的有界认证信封
 #[derive(Clone, PartialEq, Eq)]
 pub struct ClientAuthenticationRequest {
-    settings: Option<crate::settings::RequestSettings>,
+    settings: Option<crate::routing::request_settings::RequestSettings>,
     authorization: Arc<str>,
     native_bearer: Option<PlaintextClientApiKey>,
 }
 
 impl ClientAuthenticationRequest {
     #[must_use]
-    pub fn with_settings(mut self, settings: crate::settings::RequestSettings) -> Self {
+    pub fn with_settings(
+        mut self,
+        settings: crate::routing::request_settings::RequestSettings,
+    ) -> Self {
         self.settings = Some(settings);
         self
     }
 
     #[must_use]
-    pub fn settings(&self) -> Option<&crate::settings::RequestSettings> {
+    pub fn settings(&self) -> Option<&crate::routing::request_settings::RequestSettings> {
         self.settings.as_ref()
     }
 
-    pub(crate) fn take_settings(&mut self) -> Option<crate::settings::RequestSettings> {
+    pub(crate) fn take_settings(
+        &mut self,
+    ) -> Option<crate::routing::request_settings::RequestSettings> {
         self.settings.take()
     }
 

backend/crates/gateway-core/src/engine/execution/admission.rs

@@ -0,0 +1,71 @@
+//! 请求准入租约及费用结算的释放边界
+
+use super::failure::record_resource_failure;
+use crate::{
+    engine::{
+        ModelRequestId,
+        admission::ClientAdmissionPort,
+        budget::{ClientBudgetCharge, ClientBudgetPort},
+    },
+    lifecycle::LeaseGuard,
+    policy::ClientApiKeyId,
+};
+use std::sync::Arc;
+
+pub(super) struct AdmissionLease {
+    pub(super) renewal: Option<Box<dyn LeaseGuard>>,
+    pub(super) armed: bool,
+    pub(super) port: Arc<dyn ClientAdmissionPort>,
+    pub(super) diagnostics: Arc<dyn crate::diagnostics::OperationalDiagnostics>,
+    pub(super) client_api_key_id: ClientApiKeyId,
+    pub(super) model_request_id: ModelRequestId,
+}
+
+pub(super) async fn settle_budget(
+    port: &dyn ClientBudgetPort,
+    charge: ClientBudgetCharge,
+    diagnostics: &dyn crate::diagnostics::OperationalDiagnostics,
+) {
+    let request_id = charge.request_id.clone();
+    if let Err(error) = port.settle(charge).await {
+        record_resource_failure(
+            diagnostics,
+            &request_id,
+            "settle_client_budget",
+            "Client budget settlement failed; storage will retry on the next request",
+            error,
+        )
+        .await;
+    }
+}
+
+impl AdmissionLease {
+    pub(super) async fn release(mut self) {
+        self.renewal.take();
+        if let Err(error) = self
+            .port
+            .release(&self.client_api_key_id, &self.model_request_id)
+            .await
+        {
+            record_resource_failure(
+                self.diagnostics.as_ref(),
+                &self.model_request_id,
+                "release_client_admission",
+                "Client admission release failed; lease TTL remains active",
+                error,
+            )
+            .await;
+        }
+        self.armed = false;
+    }
+}
+
+impl Drop for AdmissionLease {
+    fn drop(&mut self) {
+        self.renewal.take();
+        if self.armed {
+            self.port
+                .abandon(&self.client_api_key_id, &self.model_request_id);
+        }
+    }
+}

backend/crates/gateway-core/src/engine/execution/contract.rs

@@ -0,0 +1,480 @@
+//! 执行请求、认证事实与对外会话端口
+
+use crate::{
+    engine::{
+        CoordinatedEvent, EngineError, ModelRequestId,
+        authentication::ClientAuthenticationRequest,
+        continuation::PreviousResponseId,
+        extensions::ExtensionCallScope,
+        middleware::{
+            FrozenMiddlewarePlan, MiddlewareAuthority, MiddlewareContext, MiddlewareTarget,
+        },
+        nested::ExecutionEffects,
+    },
+    error::{GatewayError, GatewayErrorKind},
+    event::{ProviderEvent, ProviderResponseHeader},
+    identity::ProviderKind,
+    lifecycle::{CancellationToken, Deadline},
+    operation::Operation,
+    policy::{ClientApiKeyId, ClientPolicy},
+    routing::{
+        ProviderCatalogUnavailable, PublicModelDescriptor, PublicModelId, RuntimeSnapshot,
+        UpstreamModelId, request_settings::RequestSettings,
+    },
+};
+use futures::future::BoxFuture;
+use std::{
+    fmt,
+    net::IpAddr,
+    sync::Arc,
+    time::{Duration, SystemTime},
+};
+use uuid::Uuid;
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq)]
+pub enum ClientTransport {
+    HttpJson,
+    HttpSse,
+    WebSocket,
+    InternalProbe,
+    InternalPlugin,
+}
+
+impl ClientTransport {
+    #[must_use]
+    pub const fn as_str(self) -> &'static str {
+        match self {
+            Self::HttpJson => "http_json",
+            Self::HttpSse => "http_sse",
+            Self::WebSocket => "websocket",
+            Self::InternalProbe => "internal",
+            Self::InternalPlugin => "internal_plugin",
+        }
+    }
+}
+
+/// API 解码后交给 Core 的稳定请求元数据
+#[derive(Debug, Clone, PartialEq, Eq)]
+pub struct ExecutionRequestMetadata {
+    pub protocol: String,
+    pub endpoint: String,
+    pub transport: ClientTransport,
+    pub stream: bool,
+    pub client_ip: Option<IpAddr>,
+    pub user_agent: Option<String>,
+    pub previous_response_id: Option<PreviousResponseId>,
+}
+
+#[derive(Clone)]
+pub struct AuthenticatedClient {
+    pub(super) settings: Option<RequestSettings>,
+    pub(super) snapshot: Arc<RuntimeSnapshot>,
+    pub(super) policy: ClientPolicy,
+    pub(super) authentication: Option<ClientAuthenticationRequest>,
+}
+
+impl AuthenticatedClient {
+    pub(super) fn execution_timeout(&self) -> Option<Duration> {
+        self.settings
+            .as_ref()
+            .and_then(|settings| settings.execution_timeout(self.policy.key_id()))
+    }
+
+    #[must_use]
+    pub fn request_settings(&self) -> Option<&RequestSettings> {
+        self.settings.as_ref()
+    }
+
+    /// 后续仍须重新认证；这里只更新下一次认证使用的冻结配置
+    #[must_use]
+    pub fn with_request_settings(mut self, settings: RequestSettings) -> Self {
+        self.settings = Some(settings);
+        self
+    }
+
+    #[must_use]
+    pub const fn snapshot(&self) -> &Arc<RuntimeSnapshot> {
+        &self.snapshot
+    }
+
+    #[must_use]
+    pub const fn policy(&self) -> &ClientPolicy {
+        &self.policy
+    }
+}
+
+impl fmt::Debug for AuthenticatedClient {
+    fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+        formatter
+            .debug_struct("AuthenticatedClient")
+            .field("key_id", &self.policy.key_id())
+            .field("revision", &self.snapshot.revision())
+            .finish_non_exhaustive()
+    }
+}
+
+#[derive(Debug, Clone, Copy, PartialEq, Eq, thiserror::Error)]
+pub enum ClientAuthenticationError {
+    #[error("client API key is invalid")]
+    InvalidKey,
+    #[error("runtime snapshot is unavailable")]
+    SnapshotUnavailable,
+    #[error("frontend authentication provider is unavailable")]
+    ProviderUnavailable,
+}
+
+pub struct StartExecution {
+    pub client: AuthenticatedClient,
+    pub public_model: PublicModelId,
+    pub operation: Operation,
+    pub metadata: ExecutionRequestMetadata,
+}
+
+/// 入口中间件开始前冻结的根请求身份与生命周期
+///
+/// 中间件短路时本值直接释放；调用 `next` 后必须原样交回
+/// [`ExecutionService::start_prepared`]，不能重新认证为另一个 Key
+pub struct PreparedRootExecution {
+    pub(super) extension_scope: ExtensionCallScope,
+    pub(super) response_control: crate::engine::response_control::ResponseControl,
+    pub(super) client: AuthenticatedClient,
+    pub(super) request_id: ModelRequestId,
+    pub(super) started_at: SystemTime,
+    pub(super) deadline_at: Deadline,
+    pub(super) cancellation: CancellationToken,
+    pub(super) execution_effects: Arc<ExecutionEffects>,
+    pub(super) execution_effects_baseline: usize,
+}
+
+impl PreparedRootExecution {
+    #[must_use]
+    pub fn request_settings(&self) -> RequestSettings {
+        self.client
+            .settings
+            .clone()
+            .unwrap_or_else(|| RequestSettings::new(self.client.snapshot.clone()))
+            .with_execution(
+                &self.client.policy,
+                self.deadline_at
+                    .at()
+                    .map(|at| duration_ms(at.duration_since(self.started_at).unwrap_or_default())),
+            )
+    }
+
+    /// 入口延续发起实例集合；保留已经交给调用方的取消句柄
+    #[must_use]
+    pub fn with_extension_scope(mut self, scope: ExtensionCallScope) -> Self {
+        self.extension_scope = scope;
+        self
+    }
+
+    #[must_use]
+    pub fn extension_scope(&self) -> ExtensionCallScope {
+        self.extension_scope.clone()
+    }
+
+    /// 设置已经在改写时完成编译；此处只将同一份结果应用到冻结身份和生命周期
+    pub fn apply_settings(&mut self, settings: &RequestSettings) -> Result<(), GatewayError> {
+        let deadline = settings
+            .execution_deadline(self.client.policy.key_id(), self.started_at)
+            .map_err(|_| {
+                GatewayError::new(
+                    GatewayErrorKind::InvalidRequest,
+                    "execution settings are invalid",
+                )
+            })?;
+        self.client.policy = settings.apply_policy(self.client.policy.clone());
+        self.client.snapshot = settings.snapshot();
+        self.client.settings = Some(settings.clone());
+        self.deadline_at = deadline;
+        Ok(())
+    }
+
+    #[must_use]
+    pub fn response_control(&self) -> crate::engine::response_control::ResponseControl {
+        self.response_control.clone()
+    }
+
+    pub(super) fn new(client: AuthenticatedClient) -> Result<Self, GatewayError> {
+        let started_at = SystemTime::now();
+        let deadline_at = Deadline::from_timeout(started_at, client.execution_timeout())
+            .ok_or_else(|| {
+                GatewayError::new(GatewayErrorKind::Internal, "system clock is invalid")
+            })?;
+        let execution_effects = Arc::new(ExecutionEffects::default());
+        let execution_effects_baseline = execution_effects.epoch();
+        Ok(Self {
+            extension_scope: ExtensionCallScope::default(),
+            client,
+            response_control: crate::engine::response_control::ResponseControl::default(),
+            request_id: new_request_id()?,
+            started_at,
+            deadline_at,
+            cancellation: CancellationToken::new(),
+            execution_effects,
+            execution_effects_baseline,
+        })
+    }
+
+    #[must_use]
+    pub const fn request_id(&self) -> &ModelRequestId {
+        &self.request_id
+    }
+
+    #[must_use]
+    pub const fn client(&self) -> &AuthenticatedClient {
+        &self.client
+    }
+
+    #[must_use]
+    pub const fn started_at(&self) -> SystemTime {
+        self.started_at
+    }
+
+    #[must_use]
+    pub const fn deadline_at(&self) -> Deadline {
+        self.deadline_at
+    }
+
+    #[must_use]
+    pub fn cancellation(&self) -> CancellationToken {
+        self.cancellation.clone()
+    }
+
+    /// 以冻结的入口身份补齐 Runtime-only binding 事实；调用位置事实由 API/Core 提供
+    #[must_use]
+    pub fn middleware_context(
+        &self,
+        mut target: MiddlewareTarget,
+        extension_scope: ExtensionCallScope,
+    ) -> MiddlewareContext {
+        target.request_id = self.request_id.clone();
+        let account_group_ids = self
+            .client
+            .policy
+            .account_scope()
+            .routing_snapshot()
+            .groups_snapshot()
+            .iter()
+            .map(|group| group.id().clone())
+            .collect::<Vec<_>>();
+        MiddlewareContext::new(
+            target,
+            MiddlewareAuthority {
+                client_key_id: self.client.policy.key_id().clone(),
+                account_group_ids: Arc::from(account_group_ids),
+                cancellation: self.cancellation.clone(),
+                deadline: self.deadline_at,
+                extension_scope,
+                execution_effects: Some(Arc::clone(&self.execution_effects)),
+            },
+        )
+    }
+}
+
+/// 入口中间件调用 `next` 后交给 Core 的已解码执行输入
+pub struct PreparedExecutionRequest {
+    pub public_model: PublicModelId,
+    pub operation: Operation,
+    pub metadata: ExecutionRequestMetadata,
+}
+
+/// 启动一个由协议 adapter 明确绑定到 Provider 自有端点的请求
+pub struct StartProviderExecution {
+    pub client: AuthenticatedClient,
+    pub provider: ProviderKind,
+    /// Provider 自有模型端点可固定上游模型；非模型端点保持 `None`
+    pub upstream_model: Option<UpstreamModelId>,
+    pub operation: Operation,
+    pub metadata: ExecutionRequestMetadata,
+}
+
+pub struct StartedExecution {
+    pub request_id: ModelRequestId,
+    pub created_at: SystemTime,
+    pub stream: bool,
+    pub session: Box<dyn ExecutionSession>,
+}
+
+pub trait ExecutionSession: Send {
+    fn trace(&self) -> crate::diagnostics::TraceContext {
+        crate::diagnostics::TraceContext::default()
+    }
+    fn next_event(&mut self) -> BoxFuture<'_, Result<Option<CoordinatedEvent>, EngineError>>;
+    fn collect_uncommitted(&mut self) -> BoxFuture<'_, Result<Vec<ProviderEvent>, EngineError>>;
+    fn response_headers(&self) -> &[ProviderResponseHeader];
+    fn response_status_code(&self) -> Option<u16> {
+        None
+    }
+    fn discard_pending_delivery(&mut self) -> Result<(), EngineError> {
+        Err(EngineError::InvalidDeliveryState)
+    }
+    fn commit_downstream(
+        &mut self,
+        client_status_code: Option<u16>,
+    ) -> BoxFuture<'_, Result<(), EngineError>>;
+    fn record_client_status(
+        &mut self,
+        client_status_code: u16,
+    ) -> BoxFuture<'_, Result<(), EngineError>>;
+    /// 执行已终结且结算、准入释放均已返回，协议层才可以放弃清理责任
+    /// 结算失败仍由 Store 保留费用重试，释放失败仍按租约 TTL 收敛
+    fn is_finalized(&self) -> bool;
+    fn cancel(&self);
+    /// 将会话交给宿主持续驱动清理；取消请求事件的等待不会丢弃已开始的结算
+    fn detach_finalize(self: Box<Self>) -> BoxFuture<'static, ()>;
+}
+
+pub trait ExecutionService: Send + Sync {
+    /// 未就绪时管理与健康接口仍可运行，模型认证按既有路径报告不可用
+    fn request_settings(&self) -> Option<RequestSettings> {
+        None
+    }
+
+    fn authenticate(
+        &self,
+        plaintext: &str,
+    ) -> Result<AuthenticatedClient, ClientAuthenticationError>;
+    fn authenticate_request(
+        &self,
+        request: ClientAuthenticationRequest,
+    ) -> BoxFuture<'_, Result<AuthenticatedClient, ClientAuthenticationError>> {
+        Box::pin(async move {
+            let plaintext = request
+                .native_bearer()
+                .ok_or(ClientAuthenticationError::InvalidKey)?;
+            self.authenticate(plaintext.expose_for_auth())
+        })
+    }
+    /// 只验证入口认证信封，不记录 Key 使用事实或执行推理准入
+    fn verify_request(
+        &self,
+        _request: ClientAuthenticationRequest,
+    ) -> BoxFuture<'_, Result<AuthenticatedClient, ClientAuthenticationError>> {
+        Box::pin(async { Err(ClientAuthenticationError::InvalidKey) })
+    }
+    fn public_models(&self, client: &AuthenticatedClient) -> Vec<PublicModelId>;
+    fn client_model_catalog<'a>(
+        &'a self,
+        _client: &'a AuthenticatedClient,
+        _protocol: &'a str,
+        _client_version: &'a str,
+    ) -> BoxFuture<'a, Result<Vec<PublicModelDescriptor>, ProviderCatalogUnavailable>> {
+        Box::pin(async { Err(ProviderCatalogUnavailable) })
+    }
+    fn contains_public_model(&self, client: &AuthenticatedClient, model: &PublicModelId) -> bool;
+
+    /// 重新鉴权并冻结本次根请求身份、请求 ID、deadline 与取消域
+    fn prepare_execution(
+        &self,
+        client: AuthenticatedClient,
+    ) -> BoxFuture<'_, Result<PreparedRootExecution, GatewayError>> {
+        Box::pin(async move { PreparedRootExecution::new(client) })
+    }
+
+    /// 由已经验证管理会话与插件 models 域的宿主入口选择 Key
+    /// 不向页面提供明文，也不跳过普通请求的扩展、准入和计量
+    fn prepare_plugin_execution(
+        &self,
+        _client_key_id: &ClientApiKeyId,
+    ) -> BoxFuture<'_, Result<PreparedRootExecution, GatewayError>> {
+        Box::pin(async {
+            Err(GatewayError::new(
+                GatewayErrorKind::Unauthorized,
+                "plugin model entry is unavailable",
+            ))
+        })
+    }
+
+    /// 为已经通过 `verify_request` 的只读入口建立中间件生命周期，不重复写 Key 使用事实
+    fn prepare_verified_execution(
+        &self,
+        client: AuthenticatedClient,
+    ) -> Result<PreparedRootExecution, GatewayError> {
+        PreparedRootExecution::new(client)
+    }
+
+    /// 解析与准备阶段冻结的发布代次一致的中间件计划
+    fn middleware_plan(&self, _prepared: &PreparedRootExecution) -> Option<FrozenMiddlewarePlan> {
+        None
+    }
+
+    /// 消费一次已准备身份并进入原有 Core 路由、准入、attempt 与结算路径
+    fn start_prepared(
+        &self,
+        prepared: PreparedRootExecution,
+        request: PreparedExecutionRequest,
+    ) -> BoxFuture<'_, Result<StartedExecution, GatewayError>> {
+        Box::pin(async move {
+            self.start(StartExecution {
+                client: prepared.client,
+                public_model: request.public_model,
+                operation: request.operation,
+                metadata: request.metadata,
+            })
+            .await
+        })
+    }
+
+    /// 消费一次已准备身份，并进入 Provider 自有端点的原有准入与结算路径
+    fn start_prepared_provider_endpoint(
+        &self,
+        prepared: PreparedRootExecution,
+        provider: ProviderKind,
+        upstream_model: Option<UpstreamModelId>,
+        operation: Operation,
+        metadata: ExecutionRequestMetadata,
+    ) -> BoxFuture<'_, Result<StartedExecution, GatewayError>> {
+        Box::pin(async move {
+            self.start_provider_endpoint(StartProviderExecution {
+                client: prepared.client,
+                provider,
+                upstream_model,
+                operation,
+                metadata,
+            })
+            .await
+        })
+    }
+
+    fn start(
+        &self,
+        request: StartExecution,
+    ) -> BoxFuture<'_, Result<StartedExecution, GatewayError>>;
+    fn start_provider_endpoint(
+        &self,
+        request: StartProviderExecution,
+    ) -> BoxFuture<'_, Result<StartedExecution, GatewayError>>;
+
+    /// 返回 Provider 注册表中首个声明 Live sideband 能力的网关。
+    /// `None` 表示当前组合不含语音 sideband；协议 adapter 据此回退到稳定 501。
+    fn live_gateway(&self) -> Option<Arc<dyn crate::live::LiveGateway>> {
+        None
+    }
+}
+
+/// 只验证 Client API Key 并返回稳定 Key ID，不产生请求使用事实
+///
+/// 管理侧登录等非数据面场景必须使用该端口，避免把认证本身误记为一次 Key 使用
+pub trait ClientKeyVerifier: Send + Sync {
+    fn verify_client_key(
+        &self,
+        plaintext: &str,
+    ) -> Result<ClientApiKeyId, ClientAuthenticationError>;
+}
+
+/// 成功认证后的 API Key 使用事实接收器
+///
+/// 认证仍是同步快照读取；实现必须自行异步、去重地持久化，不得阻塞客户端请求
+pub trait ClientApiKeyUsageSink: Send + Sync {
+    fn record_used(&self, key_id: &ClientApiKeyId);
+}
+
+pub(super) fn duration_ms(duration: Duration) -> u64 {
+    u64::try_from(duration.as_millis()).unwrap_or(u64::MAX)
+}
+
+pub(super) fn new_request_id() -> Result<ModelRequestId, GatewayError> {
+    ModelRequestId::new(format!("req_{}", Uuid::now_v7().simple()))
+        .map_err(|_| GatewayError::new(GatewayErrorKind::Internal, "failed to allocate request ID"))
+}

backend/crates/gateway-core/src/engine/execution/failure.rs

@@ -0,0 +1,100 @@
+//! 执行边界的错误分类与资源释放诊断
+
+use super::contract::ClientAuthenticationError;
+use crate::{
+    engine::{EngineError, ModelRequestId},
+    error::{GatewayError, GatewayErrorKind},
+};
+
+pub(super) fn authentication_gateway_error(error: ClientAuthenticationError) -> GatewayError {
+    match error {
+        ClientAuthenticationError::InvalidKey => {
+            GatewayError::new(GatewayErrorKind::Unauthorized, "client API key is invalid")
+        }
+        ClientAuthenticationError::SnapshotUnavailable => GatewayError::new(
+            GatewayErrorKind::Internal,
+            "runtime snapshot is unavailable",
+        ),
+        ClientAuthenticationError::ProviderUnavailable => GatewayError::new(
+            GatewayErrorKind::Internal,
+            "frontend authentication provider is unavailable",
+        ),
+    }
+}
+
+pub(super) fn map_routing_error(error: crate::validation::RoutingError) -> GatewayError {
+    match error {
+        crate::validation::RoutingError::ModelNotFound {
+            model,
+            mapped_model,
+        } => GatewayError::new(
+            GatewayErrorKind::ModelNotFound,
+            if model == mapped_model {
+                "the requested model was not found in the provider catalogs available to this API key; check the model name"
+            } else {
+                "the requested model maps to an upstream model that was not found in the provider catalogs available to this API key; check the configured model mapping"
+            },
+        ),
+        crate::validation::RoutingError::NoCapableProvider { .. }
+        | crate::validation::RoutingError::NoCapableProviderEndpoint { .. }
+        | crate::validation::RoutingError::EmptyAccountScope => GatewayError::new(
+            GatewayErrorKind::NoAvailableProvider,
+            "no provider can execute this request",
+        ),
+        crate::validation::RoutingError::UnsupportedProviderEndpoint { .. } => GatewayError::new(
+            GatewayErrorKind::Unsupported,
+            "the selected provider does not support this operation",
+        ),
+        _ => GatewayError::new(
+            GatewayErrorKind::Internal,
+            "runtime routing configuration is invalid",
+        ),
+    }
+}
+
+pub fn gateway_error_from_engine(error: &EngineError) -> GatewayError {
+    match error {
+        EngineError::Cancelled => {
+            GatewayError::new(GatewayErrorKind::Cancelled, "request was cancelled")
+        }
+        EngineError::Deadline => {
+            GatewayError::new(GatewayErrorKind::Timeout, "request deadline elapsed")
+        }
+        EngineError::Provider(provider) => GatewayError::from_provider(provider),
+        EngineError::EmptyRoutingPlan | EngineError::ProviderNotRegistered { .. } => {
+            GatewayError::new(
+                GatewayErrorKind::NoAvailableProvider,
+                "no provider is available",
+            )
+        }
+        EngineError::Store(source) => {
+            GatewayError::new(GatewayErrorKind::Internal, "request execution failed")
+                .with_source(source.clone())
+        }
+        _ => GatewayError::new(GatewayErrorKind::Internal, "request execution failed"),
+    }
+}
+
+pub(super) async fn record_resource_failure(
+    diagnostics: &dyn crate::diagnostics::OperationalDiagnostics,
+    request_id: &ModelRequestId,
+    operation: &'static str,
+    message: &'static str,
+    source: impl Into<crate::error::ErrorSource>,
+) {
+    let mut failure = crate::diagnostics::OperationalFailure::new(
+        "core",
+        operation,
+        "resource_unavailable",
+        message,
+    );
+    failure.correlation_id = Some(request_id.as_str().to_owned());
+    failure.details = crate::error::ErrorDetails::capture(Some(&source.into()), None, false);
+    if diagnostics.record_failure(failure).await.is_err() {
+        tracing::warn!(
+            request_id = request_id.as_str(),
+            operation,
+            "resource diagnostic could not be recorded"
+        );
+    }
+}

backend/crates/gateway-core/src/engine/execution/mod.rs

@@ -0,0 +1,24 @@
+//! 数据面执行入口与请求、作用域、会话的单一所有权
+
+mod admission;
+mod contract;
+mod failure;
+pub use contract::{
+    AuthenticatedClient, ClientApiKeyUsageSink, ClientAuthenticationError, ClientKeyVerifier,
+    ClientTransport, ExecutionRequestMetadata, ExecutionService, ExecutionSession,
+    PreparedExecutionRequest, PreparedRootExecution, StartExecution, StartProviderExecution,
+    StartedExecution,
+};
+mod scope;
+pub use scope::BoundModelExecutionContext;
+mod service;
+pub use failure::gateway_error_from_engine;
+pub use service::DefaultExecutionService;
+mod probe_store;
+mod session;
+
+use std::time::Duration;
+const DIAGNOSTIC_TIMEOUT: Duration = Duration::from_secs(10 * 60);
+const COORDINATION_TIMEOUT: Duration = Duration::from_millis(100);
+const MAX_NESTED_EXECUTIONS: usize = 16;
+const MAX_CONCURRENT_NESTED_EXECUTIONS: usize = 4;

backend/crates/gateway-core/src/engine/execution/probe_store.rs

@@ -0,0 +1,64 @@
+//! 诊断探测使用的非持久执行记录与观测事实
+
+use crate::{
+    engine::{
+        AttemptRecord, ExecutionStore, IntermediateFailure, ModelRequestFinalization,
+        ModelRequestId, NewModelRequest, ProviderAccountId, RecoveryReport, UpstreamSendState,
+    },
+    error::StoreError,
+    identity::ProviderKind,
+    routing::UpstreamModelId,
+};
+use std::time::SystemTime;
+
+pub(super) struct ProbeObservation {
+    pub(super) provider_kind: ProviderKind,
+    pub(super) account_id: ProviderAccountId,
+    pub(super) upstream_model: UpstreamModelId,
+}
+
+pub(super) struct TransientExecutionStore;
+
+#[async_trait::async_trait]
+impl ExecutionStore for TransientExecutionStore {
+    async fn create_model_request(&self, _: NewModelRequest) -> Result<(), StoreError> {
+        Ok(())
+    }
+
+    async fn record_attempt(&self, _: AttemptRecord) -> Result<(), StoreError> {
+        Ok(())
+    }
+
+    async fn mark_send_state(
+        &self,
+        _: &ModelRequestId,
+        _: UpstreamSendState,
+    ) -> Result<(), StoreError> {
+        Ok(())
+    }
+
+    async fn mark_downstream_committed(
+        &self,
+        _: &ModelRequestId,
+        _: SystemTime,
+        _: Option<u16>,
+    ) -> Result<(), StoreError> {
+        Ok(())
+    }
+
+    async fn record_client_status(&self, _: &ModelRequestId, _: u16) -> Result<(), StoreError> {
+        Ok(())
+    }
+
+    async fn record_intermediate_failure(&self, _: IntermediateFailure) -> Result<(), StoreError> {
+        Ok(())
+    }
+
+    async fn finalize_model_request(&self, _: ModelRequestFinalization) -> Result<(), StoreError> {
+        Ok(())
+    }
+
+    async fn recover_expired(&self, _: SystemTime) -> Result<RecoveryReport, StoreError> {
+        Ok(RecoveryReport::default())
+    }
+}

backend/crates/gateway-core/src/engine/execution/scope.rs

@@ -0,0 +1,301 @@
+//! 请求嵌套图、活动权限与作用域 guard
+
+use super::admission::{AdmissionLease, settle_budget};
+use super::contract::{AuthenticatedClient, ExecutionRequestMetadata};
+use super::{MAX_CONCURRENT_NESTED_EXECUTIONS, MAX_NESTED_EXECUTIONS};
+use crate::engine::budget::{ClientBudgetCharge, ClientBudgetPort};
+use crate::{
+    concurrency::ConcurrencyWaitBudget,
+    engine::{
+        ModelRequestId, ProviderAccountId, coordinator::CoordinationExtensions,
+        extensions::ExtensionCallScope, nested::ExecutionEffects,
+    },
+    error::{GatewayError, GatewayErrorKind},
+    identity::ProviderKind,
+    lifecycle::{CancellationToken, Deadline},
+    operation::Operation,
+    policy::ClientApiKeyId,
+    routing::{
+        FrozenAccountScope, PublicModelId, UpstreamModelId, request_settings::RequestSettings,
+    },
+};
+use std::{
+    collections::BTreeMap,
+    fmt,
+    sync::{
+        Arc, Mutex, Weak,
+        atomic::{AtomicUsize, Ordering},
+    },
+    time::SystemTime,
+};
+
+pub(super) enum ExecutionTarget {
+    Model(PublicModelId),
+    ProviderEndpoint {
+        provider: ProviderKind,
+        upstream_model: Option<UpstreamModelId>,
+    },
+}
+
+impl ExecutionTarget {
+    pub(super) fn public_model(&self) -> Option<&PublicModelId> {
+        match self {
+            Self::Model(model) => Some(model),
+            Self::ProviderEndpoint { .. } => None,
+        }
+    }
+
+    pub(super) fn into_public_model(self) -> Option<PublicModelId> {
+        match self {
+            Self::Model(model) => Some(model),
+            Self::ProviderEndpoint { .. } => None,
+        }
+    }
+}
+
+pub(super) struct PendingStartExecution {
+    pub(super) client: AuthenticatedClient,
+    pub(super) target: ExecutionTarget,
+    pub(super) operation: Operation,
+    pub(super) metadata: ExecutionRequestMetadata,
+}
+
+pub(super) struct AuthorizedExecution {
+    pub(super) response_control: Option<crate::engine::response_control::ResponseControl>,
+    pub(super) account_scope: Arc<FrozenAccountScope>,
+    pub(super) deadline_at: Deadline,
+    pub(super) cancellation: CancellationToken,
+    pub(super) extension_scope: ExtensionCallScope,
+    pub(super) required_provider: Option<ProviderKind>,
+    pub(super) required_account: Option<ProviderAccountId>,
+    pub(super) nested: Option<NestedExecutionFacts>,
+    pub(super) bound: Option<BoundExecutionFacts>,
+    pub(super) graph: Option<Arc<NestedExecutionGraph>>,
+    pub(super) execution_effects_baseline: usize,
+    pub(super) nested_permit: Option<NestedExecutionPermit>,
+}
+
+pub(super) struct ExecutionStartGuard {
+    pub(super) admission: ExecutionAdmission,
+    pub(super) active_request: Option<ActiveRequestLease>,
+    pub(super) concurrency_wait_budget: ConcurrencyWaitBudget,
+    pub(super) admission_decision_ms: Option<u64>,
+}
+
+pub(super) struct PreparedExecutionStart {
+    pub(super) request_id: ModelRequestId,
+    pub(super) started_at: SystemTime,
+    pub(super) plan: crate::routing::RoutingPlan,
+    pub(super) extensions: CoordinationExtensions,
+    pub(super) authorization: AuthorizedExecution,
+    pub(super) guard: ExecutionStartGuard,
+}
+
+pub(super) struct NestedExecutionFacts {
+    pub(super) parent_request_id: ModelRequestId,
+    pub(super) initiating_plugin_instance_id: String,
+}
+
+pub(super) struct BoundExecutionFacts {
+    pub(super) initiating_plugin_instance_id: String,
+}
+
+pub(super) struct NestedExecutionGraph {
+    pub(super) total_started: AtomicUsize,
+    pub(super) active: AtomicUsize,
+    pub(super) effects: Arc<ExecutionEffects>,
+}
+
+impl NestedExecutionGraph {
+    pub(super) fn new(effects: Arc<ExecutionEffects>) -> Self {
+        Self {
+            total_started: AtomicUsize::new(0),
+            active: AtomicUsize::new(0),
+            effects,
+        }
+    }
+
+    pub(super) fn acquire(self: &Arc<Self>) -> Result<NestedExecutionPermit, GatewayError> {
+        if self
+            .active
+            .fetch_update(Ordering::AcqRel, Ordering::Acquire, |active| {
+                (active < MAX_CONCURRENT_NESTED_EXECUTIONS).then_some(active + 1)
+            })
+            .is_err()
+        {
+            return Err(GatewayError::new(
+                GatewayErrorKind::ConcurrencyQueueFull,
+                "nested model execution concurrency is exhausted",
+            ));
+        }
+        if self
+            .total_started
+            .fetch_update(Ordering::AcqRel, Ordering::Acquire, |total| {
+                (total < MAX_NESTED_EXECUTIONS).then_some(total + 1)
+            })
+            .is_err()
+        {
+            self.active.fetch_sub(1, Ordering::AcqRel);
+            return Err(GatewayError::new(
+                GatewayErrorKind::PolicyDenied,
+                "nested model execution budget is exhausted",
+            ));
+        }
+        Ok(NestedExecutionPermit {
+            graph: Arc::clone(self),
+            armed: true,
+        })
+    }
+}
+
+pub(super) struct NestedExecutionPermit {
+    pub(super) graph: Arc<NestedExecutionGraph>,
+    pub(super) armed: bool,
+}
+
+impl NestedExecutionPermit {
+    pub(super) fn release(mut self) {
+        if self.armed {
+            self.graph.active.fetch_sub(1, Ordering::AcqRel);
+            self.armed = false;
+        }
+    }
+}
+
+impl Drop for NestedExecutionPermit {
+    fn drop(&mut self) {
+        if self.armed {
+            self.graph.active.fetch_sub(1, Ordering::AcqRel);
+        }
+    }
+}
+
+pub(super) struct ActiveRequestAuthority {
+    pub(super) client: AuthenticatedClient,
+    pub(super) account_scope: Arc<FrozenAccountScope>,
+    pub(super) deadline_at: Deadline,
+    pub(super) cancellation: CancellationToken,
+    pub(super) extension_scope: ExtensionCallScope,
+    pub(super) graph: Arc<NestedExecutionGraph>,
+}
+
+pub(super) type ActiveRequestRegistry =
+    Arc<Mutex<BTreeMap<ModelRequestId, Weak<ActiveRequestAuthority>>>>;
+
+pub(super) struct ActiveRequestLease {
+    pub(super) registry: ActiveRequestRegistry,
+    pub(super) request_id: ModelRequestId,
+    pub(super) authority: Arc<ActiveRequestAuthority>,
+}
+
+/// 管理页或 CLI 一次调用持有的显式模型执行身份
+///
+/// 结构不暴露 Key 明文；Runtime 只能把它原样交回 Core 发起模型请求
+#[derive(Clone)]
+pub struct BoundModelExecutionContext {
+    pub(super) authority: Arc<BoundModelExecutionAuthority>,
+}
+
+impl BoundModelExecutionContext {
+    #[must_use]
+    pub fn request_settings(&self) -> Option<&RequestSettings> {
+        self.authority.client.request_settings()
+    }
+}
+
+pub(super) struct BoundModelExecutionAuthority {
+    pub(super) client: AuthenticatedClient,
+    pub(super) account_scope: Arc<FrozenAccountScope>,
+    pub(super) cancellation: CancellationToken,
+    pub(super) extension_scope: ExtensionCallScope,
+    pub(super) initiating_plugin_instance_id: String,
+}
+
+impl fmt::Debug for BoundModelExecutionContext {
+    fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+        formatter
+            .debug_struct("BoundModelExecutionContext")
+            .field("key_id", &self.authority.client.policy.key_id())
+            .field(
+                "initiating_plugin_instance_id",
+                &self.authority.initiating_plugin_instance_id,
+            )
+            .finish_non_exhaustive()
+    }
+}
+
+impl Drop for BoundModelExecutionAuthority {
+    fn drop(&mut self) {
+        self.cancellation.cancel();
+    }
+}
+
+impl ActiveRequestLease {
+    pub(super) fn release(self) {
+        drop(self);
+    }
+}
+
+impl Drop for ActiveRequestLease {
+    fn drop(&mut self) {
+        self.authority.cancellation.cancel();
+        let mut active = self
+            .registry
+            .lock()
+            .unwrap_or_else(std::sync::PoisonError::into_inner);
+        if active
+            .get(&self.request_id)
+            .and_then(Weak::upgrade)
+            .is_some_and(|authority| Arc::ptr_eq(&authority, &self.authority))
+        {
+            active.remove(&self.request_id);
+        }
+    }
+}
+
+impl ExecutionStartGuard {
+    pub(super) async fn release_failed(
+        self,
+        budget: Option<&dyn ClientBudgetPort>,
+        request_id: ModelRequestId,
+        key_id: ClientApiKeyId,
+        diagnostics: &dyn crate::diagnostics::OperationalDiagnostics,
+    ) {
+        if let Some(active_request) = self.active_request {
+            active_request.release();
+        }
+        if let Some(budget) = budget {
+            settle_budget(
+                budget,
+                ClientBudgetCharge {
+                    key_id,
+                    request_id,
+                    amount_usd: crate::metering::Decimal::ZERO,
+                    completed_at: SystemTime::now(),
+                },
+                diagnostics,
+            )
+            .await;
+        }
+        self.admission.release().await;
+    }
+}
+
+pub(super) enum ExecutionAdmission {
+    Client(AdmissionLease),
+    Nested(NestedExecutionPermit),
+    Bound(AdmissionLease, NestedExecutionPermit),
+}
+
+impl ExecutionAdmission {
+    pub(super) async fn release(self) {
+        match self {
+            Self::Client(admission) => admission.release().await,
+            Self::Nested(permit) => permit.release(),
+            Self::Bound(admission, permit) => {
+                admission.release().await;
+                permit.release();
+            }
+        }
+    }
+}

backend/crates/gateway-core/src/engine/execution/session.rs

@@ -0,0 +1,213 @@
+//! 执行会话的准入释放、费用结算与终态收敛
+
+use super::COORDINATION_TIMEOUT;
+use super::admission::settle_budget;
+use super::contract::ExecutionSession;
+use super::scope::{ActiveRequestLease, ExecutionAdmission};
+use crate::{
+    engine::{
+        CoordinatedEvent, EngineError, ExecutionStore,
+        budget::ClientBudgetPort,
+        continuation::{NativeContinuationPin, NativeContinuationPort},
+        coordinator::ResponseExecutionSession,
+    },
+    event::{ProviderEvent, ProviderResponseHeader},
+    operation::ProviderSessionState,
+};
+use futures::{FutureExt as _, future::BoxFuture, pin_mut, select_biased};
+use futures_timer::Delay;
+use std::sync::Arc;
+
+pub(super) struct DefaultExecutionSession {
+    pub(super) core: ResponseExecutionSession<dyn ExecutionStore>,
+    pub(super) admission: Option<ExecutionAdmission>,
+    pub(super) active_request: Option<ActiveRequestLease>,
+    pub(super) cleanup: Option<BoxFuture<'static, ()>>,
+    pub(super) continuation: Arc<dyn NativeContinuationPort>,
+    pub(super) continuation_recorded: bool,
+    pub(super) budget: Option<Arc<dyn ClientBudgetPort>>,
+    pub(super) diagnostics: Arc<dyn crate::diagnostics::OperationalDiagnostics>,
+}
+
+impl DefaultExecutionSession {
+    pub(super) fn new(
+        core: ResponseExecutionSession<dyn ExecutionStore>,
+        admission: ExecutionAdmission,
+        active_request: Option<ActiveRequestLease>,
+        continuation: Arc<dyn NativeContinuationPort>,
+        budget: Option<Arc<dyn ClientBudgetPort>>,
+        diagnostics: Arc<dyn crate::diagnostics::OperationalDiagnostics>,
+    ) -> Self {
+        Self {
+            core,
+            admission: Some(admission),
+            active_request,
+            cleanup: None,
+            continuation,
+            continuation_recorded: false,
+            budget,
+            diagnostics,
+        }
+    }
+
+    pub(super) async fn settle_if_finalized(&mut self) {
+        if self.core.is_finalized()
+            && let Some(admission) = self.admission.take()
+        {
+            if let Some(active_request) = self.active_request.take() {
+                active_request.release();
+            }
+            let budget = self.budget.take();
+            let charge = self.core.budget_charge();
+            let diagnostics = self.diagnostics.clone();
+            // 在首次 await 前把完整清理责任留在会话内
+            // 事件等待被取消后，后续 poll
+            // 或 detach 继续同一个 future，既不丢失费用，也不重启已完成的结算
+            self.cleanup = Some(Box::pin(async move {
+                if let Some(budget) = budget {
+                    settle_budget(budget.as_ref(), charge, diagnostics.as_ref()).await;
+                }
+                admission.release().await;
+            }));
+        }
+        if let Some(cleanup) = self.cleanup.as_mut() {
+            cleanup.await;
+            self.cleanup = None;
+        }
+    }
+
+    pub(super) async fn record_continuation(&mut self, state: Option<&ProviderSessionState>) {
+        if self.continuation_recorded {
+            return;
+        }
+        let Some(state) = state else {
+            return;
+        };
+        let Some(pin) = self.core.native_continuation_pin(state) else {
+            return;
+        };
+        self.continuation_recorded = true;
+        record_native_continuation(self.continuation.as_ref(), pin).await;
+    }
+
+    pub(super) async fn finalize_detached(&mut self) {
+        if let Err(error) = self.core.cancel_and_finalize().await {
+            tracing::warn!(request_id = self.core.request_id().as_str(), operation = "finalize_detached_execution", %error, "Detached execution 终态收敛失败");
+        }
+        self.settle_if_finalized().await;
+    }
+}
+
+impl Drop for DefaultExecutionSession {
+    fn drop(&mut self) {
+        self.core.cancel();
+        drop(self.active_request.take());
+    }
+}
+
+impl ExecutionSession for DefaultExecutionSession {
+    fn trace(&self) -> crate::diagnostics::TraceContext {
+        self.core.trace()
+    }
+    fn next_event(&mut self) -> BoxFuture<'_, Result<Option<CoordinatedEvent>, EngineError>> {
+        Box::pin(async move {
+            let result = self.core.next_event().await;
+            if let Ok(Some(event)) = result.as_ref() {
+                self.record_continuation(event.session_update()).await;
+            }
+            self.settle_if_finalized().await;
+            result
+        })
+    }
+
+    fn collect_uncommitted(&mut self) -> BoxFuture<'_, Result<Vec<ProviderEvent>, EngineError>> {
+        Box::pin(async move {
+            let result = self.core.collect_uncommitted().await;
+            if let Ok(events) = result.as_ref() {
+                let state = events.iter().find_map(ProviderEvent::session_update);
+                self.record_continuation(state).await;
+            }
+            self.settle_if_finalized().await;
+            result
+        })
+    }
+
+    fn response_headers(&self) -> &[ProviderResponseHeader] {
+        self.core.response_headers()
+    }
+
+    fn response_status_code(&self) -> Option<u16> {
+        self.core.response_status_code()
+    }
+
+    fn discard_pending_delivery(&mut self) -> Result<(), EngineError> {
+        self.core.discard_pending_delivery()
+    }
+
+    fn commit_downstream(
+        &mut self,
+        client_status_code: Option<u16>,
+    ) -> BoxFuture<'_, Result<(), EngineError>> {
+        Box::pin(async move {
+            let result = self.core.commit_downstream(client_status_code).await;
+            self.settle_if_finalized().await;
+            result
+        })
+    }
+
+    fn record_client_status(
+        &mut self,
+        client_status_code: u16,
+    ) -> BoxFuture<'_, Result<(), EngineError>> {
+        Box::pin(async move {
+            let result = self.core.record_client_status(client_status_code).await;
+            self.settle_if_finalized().await;
+            result
+        })
+    }
+
+    fn is_finalized(&self) -> bool {
+        self.core.is_finalized()
+            && self.admission.is_none()
+            && self.active_request.is_none()
+            && self.cleanup.is_none()
+    }
+
+    fn cancel(&self) {
+        self.core.cancel();
+    }
+
+    fn detach_finalize(mut self: Box<Self>) -> BoxFuture<'static, ()> {
+        Box::pin(async move { self.finalize_detached().await })
+    }
+}
+
+pub(super) async fn record_native_continuation(
+    continuation: &dyn NativeContinuationPort,
+    pin: NativeContinuationPin,
+) {
+    let provider = pin.provider().as_str().to_owned();
+    let account = pin.account().as_str().to_owned();
+    let record = continuation.record(pin).fuse();
+    let timeout = Delay::new(COORDINATION_TIMEOUT).fuse();
+    pin_mut!(record, timeout);
+    select_biased! {
+        result = record => {
+            if let Err(error) = result {
+                tracing::warn!(
+                    provider = %provider,
+                    account = %account,
+                    %error,
+                    "Continuation affinity 写入失败，后续请求将退化为外部续接"
+                );
+            }
+        },
+        _ = timeout => {
+            tracing::warn!(
+                provider = %provider,
+                account = %account,
+                "Continuation affinity 后台写入超时，已丢弃本次亲和记录"
+            );
+        },
+    }
+}

backend/crates/gateway-core/src/engine/extensions.rs

@@ -1,4 +1,4 @@
-//! 插件嵌套调用的递归作用域
+//! 执行扩展计划的弱引用索引、冻结解析与插件递归作用域
 
 use std::{collections::BTreeSet, sync::Arc};
 
@@ -43,3 +43,82 @@ impl ExtensionCallScope {
         })
     }
 }
+
+/// 同一代次的具体执行计划；不拥有发布状态，运行资源由对应发布集合保活
+#[derive(Debug, Default)]
+pub struct ExecutionExtensionPlans {
+    pub middleware: Option<Arc<dyn super::middleware::MiddlewarePlan>>,
+    pub upstream_adapters: Option<Arc<dyn super::upstream_adapter::UpstreamAdapterPlan>>,
+}
+
+/// 同一发布代次不能被静默替换
+#[derive(Debug, Clone, Copy, PartialEq, Eq, thiserror::Error)]
+#[error("execution extension generation is already registered")]
+pub struct ExtensionRegistrationError;
+
+/// 仅以弱引用按冻结身份解析执行计划，不延长旧代次寿命或决定何时发布
+#[derive(Clone, Default)]
+pub struct ExecutionExtensionIndex {
+    sets: Arc<
+        std::sync::RwLock<
+            std::collections::BTreeMap<
+                crate::routing::extensions::ExtensionSetId,
+                std::sync::Weak<ExecutionExtensionPlans>,
+            >,
+        >,
+    >,
+}
+
+impl ExecutionExtensionIndex {
+    pub fn register(
+        &self,
+        id: crate::routing::extensions::ExtensionSetId,
+        plans: Arc<ExecutionExtensionPlans>,
+    ) -> Result<Arc<ExecutionExtensionPlans>, ExtensionRegistrationError> {
+        let mut sets = self
+            .sets
+            .write()
+            .unwrap_or_else(std::sync::PoisonError::into_inner);
+        sets.retain(|_, plans| plans.strong_count() > 0);
+        if sets.contains_key(&id) {
+            return Err(ExtensionRegistrationError);
+        }
+        sets.insert(id, Arc::downgrade(&plans));
+        Ok(plans)
+    }
+
+    fn resolve(
+        &self,
+        generation: &crate::routing::extensions::ExtensionSetReference,
+    ) -> Option<Arc<ExecutionExtensionPlans>> {
+        self.sets
+            .read()
+            .unwrap_or_else(std::sync::PoisonError::into_inner)
+            .get(generation.id())
+            .and_then(std::sync::Weak::upgrade)
+    }
+
+    #[must_use]
+    pub fn middleware(
+        &self,
+        generation: &crate::routing::extensions::ExtensionSetReference,
+    ) -> Option<super::middleware::FrozenMiddlewarePlan> {
+        let plan = self.resolve(generation)?.middleware.clone()?;
+        Some(super::middleware::FrozenMiddlewarePlan::new(
+            plan,
+            generation.clone(),
+        ))
+    }
+
+    #[must_use]
+    pub fn upstream_adapters(
+        &self,
+        generation: &crate::routing::extensions::ExtensionSetReference,
+    ) -> Option<super::upstream_adapter::FrozenUpstreamAdapterPlan> {
+        let plan = self.resolve(generation)?.upstream_adapters.clone()?;
+        Some(super::upstream_adapter::FrozenUpstreamAdapterPlan::new(
+            plan,
+            generation.clone(),
+        ))
+    }
+}

backend/crates/gateway-core/src/engine/middleware/contract.rs

@@ -5,10 +5,11 @@
 //! Runtime/SDK 负责；身份、路由、账号租约、发送事实、重试、计量与结算仍由 Core
 //! 持有，不能通过本端口交给插件解释
 
-use std::collections::{BTreeMap, BTreeSet};
+use super::FrozenMiddlewarePlan;
+use std::collections::BTreeSet;
 use std::fmt;
 use std::num::NonZeroU32;
-use std::sync::{Arc, RwLock, Weak};
+use std::sync::Arc;
 
 use bytes::Bytes;
 use futures::future::BoxFuture;
@@ -27,8 +28,7 @@ use crate::operation::{
     CapabilityRequirements, Feature, GenerateRequest, Operation, OperationKind, ProtocolPayload,
 };
 use crate::policy::ClientApiKeyId;
-use crate::runtime::extensions::{ExtensionSetId, ExtensionSetReference};
-use crate::settings::RequestSettings;
+use crate::routing::request_settings::RequestSettings;
 
 /// 中间件在一次逻辑请求中的挂载位置
 #[derive(Debug, Clone, Copy, PartialEq, Eq, PartialOrd, Ord)]
@@ -543,7 +543,7 @@ impl fmt::Debug for MiddlewareResponseEnvelope {
 /// Core 冻结的调用事实；完整业务身份可由插件读取
 #[derive(Clone)]
 pub struct MiddlewareContext {
-    plan: Option<FrozenMiddlewarePlan>,
+    pub(super) plan: Option<FrozenMiddlewarePlan>,
     request_id: ModelRequestId,
     mount: MiddlewareMount,
     attempt_index: Option<NonZeroU32>,
@@ -799,175 +799,3 @@ impl From<ProviderError> for MiddlewareError {
 /// 绑定当前调用且只能消费一次的宿主续体
 pub type MiddlewareNext =
     crate::middleware::Next<MiddlewareRequest, MiddlewareResponse, MiddlewareError>;
-
-/// 一个发布代次冻结的统一数据面中间件计划
-pub trait MiddlewarePlan: Send + Sync + fmt::Debug {
-    fn has_service(&self) -> bool {
-        false
-    }
-    fn handle_service(
-        &self,
-        _context: crate::middleware::service::Context,
-        input: crate::middleware::service::Value,
-        next: crate::middleware::service::Next,
-    ) -> BoxFuture<
-        'static,
-        Result<crate::middleware::service::Value, crate::middleware::service::Error>,
-    > {
-        next.run(input)
-    }
-
-    fn has_websocket(&self) -> bool {
-        false
-    }
-    fn handle_websocket(
-        &self,
-        _context: crate::middleware::websocket::Context,
-        message: crate::middleware::websocket::Message,
-        next: crate::middleware::websocket::Next,
-    ) -> BoxFuture<'static, Result<Option<crate::middleware::websocket::Message>, MiddlewareError>>
-    {
-        next.run(message)
-    }
-
-    /// HTTP 入口在认证、路由与正文解析之前组合，无匹配时保留直接路径
-    fn has_http(&self) -> bool {
-        false
-    }
-
-    fn handle_http(
-        &self,
-        _context: crate::middleware::http::Context,
-        request: crate::middleware::http::Request,
-        next: crate::middleware::http::Next,
-    ) -> BoxFuture<'static, Result<crate::middleware::http::Response, MiddlewareError>> {
-        next.run(request)
-    }
-
-    fn handle(
-        &self,
-        context: MiddlewareContext,
-        request: MiddlewareRequest,
-        next: MiddlewareNext,
-    ) -> BoxFuture<'static, Result<MiddlewareResponse, MiddlewareError>>;
-}
-
-/// 调用期间同时保活发布代次与中间件计划
-#[derive(Clone)]
-pub struct FrozenMiddlewarePlan {
-    plan: Arc<dyn MiddlewarePlan>,
-    _generation: ExtensionSetReference,
-}
-
-impl FrozenMiddlewarePlan {
-    #[must_use]
-    pub fn has_service(&self) -> bool {
-        self.plan.has_service()
-    }
-    pub fn handle_service(
-        &self,
-        context: crate::middleware::service::Context,
-        input: crate::middleware::service::Value,
-        next: crate::middleware::service::Next,
-    ) -> BoxFuture<
-        'static,
-        Result<crate::middleware::service::Value, crate::middleware::service::Error>,
-    > {
-        self.plan.handle_service(context, input, next)
-    }
-
-    #[must_use]
-    pub fn has_websocket(&self) -> bool {
-        self.plan.has_websocket()
-    }
-    pub fn handle_websocket(
-        &self,
-        context: crate::middleware::websocket::Context,
-        message: crate::middleware::websocket::Message,
-        next: crate::middleware::websocket::Next,
-    ) -> BoxFuture<'static, Result<Option<crate::middleware::websocket::Message>, MiddlewareError>>
-    {
-        self.plan.handle_websocket(context, message, next)
-    }
-
-    #[must_use]
-    pub fn has_http(&self) -> bool {
-        self.plan.has_http()
-    }
-
-    pub fn handle_http(
-        &self,
-        mut context: crate::middleware::http::Context,
-        request: crate::middleware::http::Request,
-        next: crate::middleware::http::Next,
-    ) -> BoxFuture<'static, Result<crate::middleware::http::Response, MiddlewareError>> {
-        context.plan = Some(self.clone());
-        self.plan.handle_http(context, request, next)
-    }
-
-    #[must_use]
-    pub fn new(plan: Arc<dyn MiddlewarePlan>, generation: ExtensionSetReference) -> Self {
-        Self {
-            plan,
-            _generation: generation,
-        }
-    }
-
-    pub fn handle(
-        &self,
-        mut context: MiddlewareContext,
-        request: MiddlewareRequest,
-        next: MiddlewareNext,
-    ) -> BoxFuture<'static, Result<MiddlewareResponse, MiddlewareError>> {
-        context.plan = Some(self.clone());
-        self.plan.handle(context, request, next)
-    }
-}
-
-impl fmt::Debug for FrozenMiddlewarePlan {
-    fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
-        formatter
-            .debug_struct("FrozenMiddlewarePlan")
-            .finish_non_exhaustive()
-    }
-}
-
-/// 同一发布代次不能被静默替换
-#[derive(Debug, Clone, Copy, PartialEq, Eq, thiserror::Error)]
-#[error("middleware generation is already registered")]
-pub struct MiddlewareRegistrationError;
-
-/// 按请求冻结的发布集合解析中间件计划；索引不延长旧代次寿命
-#[derive(Clone, Default)]
-pub struct MiddlewareExtensionIndex {
-    sets: Arc<RwLock<BTreeMap<ExtensionSetId, Weak<dyn MiddlewarePlan>>>>,
-}
-
-impl MiddlewareExtensionIndex {
-    pub fn register(
-        &self,
-        id: ExtensionSetId,
-        plan: Arc<dyn MiddlewarePlan>,
-    ) -> Result<Arc<dyn MiddlewarePlan>, MiddlewareRegistrationError> {
-        let mut sets = self
-            .sets
-            .write()
-            .unwrap_or_else(std::sync::PoisonError::into_inner);
-        sets.retain(|_, plan| plan.strong_count() > 0);
-        if sets.contains_key(&id) {
-            return Err(MiddlewareRegistrationError);
-        }
-        sets.insert(id, Arc::downgrade(&plan));
-        Ok(plan)
-    }
-
-    #[must_use]
-    pub fn resolve(&self, generation: &ExtensionSetReference) -> Option<FrozenMiddlewarePlan> {
-        self.sets
-            .read()
-            .unwrap_or_else(std::sync::PoisonError::into_inner)
-            .get(generation.id())
-            .and_then(Weak::upgrade)
-            .map(|plan| FrozenMiddlewarePlan::new(plan, generation.clone()))
-    }
-}

backend/crates/gateway-core/src/engine/middleware/http/mod.rs

@@ -13,7 +13,7 @@ use crate::{engine::middleware::MiddlewareError, lifecycle::CancellationToken};
 pub type Body = UnsyncBoxBody<Bytes, Box<dyn std::error::Error + Send + Sync>>;
 pub type Request = http::Request<Body>;
 pub type Response = http::Response<Body>;
-pub type Next = super::Next<Request, Response, MiddlewareError>;
+pub type Next = crate::middleware::Next<Request, Response, MiddlewareError>;
 
 /// 路由前调用没有 Client Key；身份在默认认证流程中产生
 #[derive(Clone, Debug)]
@@ -29,7 +29,7 @@ pub struct Context {
 
 /// 内部子请求复用 API 总路由；端口不解释业务路径，也不经网络回环
 pub trait Dispatcher: Send + Sync {
-    fn request_settings(&self) -> Option<crate::settings::RequestSettings> {
+    fn request_settings(&self) -> Option<crate::routing::request_settings::RequestSettings> {
         None
     }
 
@@ -44,7 +44,7 @@ pub trait Dispatcher: Send + Sync {
 #[derive(Clone, Debug, Default, PartialEq, Eq)]
 pub struct Settings {
     pub timeout: Option<Duration>,
-    pub runtime: Option<crate::settings::RequestSettings>,
+    pub runtime: Option<crate::routing::request_settings::RequestSettings>,
 }
 
 #[must_use]

backend/crates/gateway-core/src/engine/middleware/http/upgrade.rs

@@ -4,7 +4,7 @@ use std::sync::{Arc, Mutex};
 
 use futures::future::{BoxFuture, Shared};
 
-use crate::{engine::middleware::MiddlewareError, middleware::websocket::Session};
+use crate::{engine::middleware::MiddlewareError, engine::middleware::websocket::Session};
 
 pub type PendingSession =
     Shared<BoxFuture<'static, Result<Arc<dyn Session>, Arc<MiddlewareError>>>>;

backend/crates/gateway-core/src/engine/middleware/mod.rs

@@ -0,0 +1,15 @@
+//! 执行层的请求、HTTP、WebSocket 与公开服务中间件合同
+
+mod contract;
+pub mod http;
+mod plan;
+pub mod service;
+pub mod websocket;
+
+pub use contract::{
+    MiddlewareAuthority, MiddlewareBody, MiddlewareCapabilityDeclaration, MiddlewareContext,
+    MiddlewareError, MiddlewareFrame, MiddlewareFrameEnvelope, MiddlewareFraming, MiddlewareHeader,
+    MiddlewareMount, MiddlewareNext, MiddlewareRequest, MiddlewareResponse,
+    MiddlewareResponseEnvelope, MiddlewareTarget,
+};
+pub use plan::{FrozenMiddlewarePlan, MiddlewarePlan};

backend/crates/gateway-core/src/engine/middleware/plan.rs

@@ -0,0 +1,152 @@
+//! 与发布代次一起冻结的协议及请求中间件计划
+
+use super::{
+    MiddlewareContext, MiddlewareError, MiddlewareNext, MiddlewareRequest, MiddlewareResponse,
+};
+use crate::routing::extensions::ExtensionSetReference;
+use futures::future::BoxFuture;
+use std::{fmt, sync::Arc};
+
+/// 一个发布代次冻结的统一数据面中间件计划
+pub trait MiddlewarePlan: Send + Sync + fmt::Debug {
+    fn has_service(&self) -> bool {
+        false
+    }
+    fn handle_service(
+        &self,
+        _context: crate::engine::middleware::service::Context,
+        input: crate::engine::middleware::service::Value,
+        next: crate::engine::middleware::service::Next,
+    ) -> BoxFuture<
+        'static,
+        Result<
+            crate::engine::middleware::service::Value,
+            crate::engine::middleware::service::Error,
+        >,
+    > {
+        next.run(input)
+    }
+
+    fn has_websocket(&self) -> bool {
+        false
+    }
+    fn handle_websocket(
+        &self,
+        _context: crate::engine::middleware::websocket::Context,
+        message: crate::engine::middleware::websocket::Message,
+        next: crate::engine::middleware::websocket::Next,
+    ) -> BoxFuture<
+        'static,
+        Result<Option<crate::engine::middleware::websocket::Message>, MiddlewareError>,
+    > {
+        next.run(message)
+    }
+
+    /// HTTP 入口在认证、路由与正文解析之前组合，无匹配时保留直接路径
+    fn has_http(&self) -> bool {
+        false
+    }
+
+    fn handle_http(
+        &self,
+        _context: crate::engine::middleware::http::Context,
+        request: crate::engine::middleware::http::Request,
+        next: crate::engine::middleware::http::Next,
+    ) -> BoxFuture<'static, Result<crate::engine::middleware::http::Response, MiddlewareError>>
+    {
+        next.run(request)
+    }
+
+    fn handle(
+        &self,
+        context: MiddlewareContext,
+        request: MiddlewareRequest,
+        next: MiddlewareNext,
+    ) -> BoxFuture<'static, Result<MiddlewareResponse, MiddlewareError>>;
+}
+
+/// 调用期间同时保活发布代次与中间件计划
+#[derive(Clone)]
+pub struct FrozenMiddlewarePlan {
+    plan: Arc<dyn MiddlewarePlan>,
+    _generation: ExtensionSetReference,
+}
+
+impl FrozenMiddlewarePlan {
+    #[must_use]
+    pub fn has_service(&self) -> bool {
+        self.plan.has_service()
+    }
+    pub fn handle_service(
+        &self,
+        context: crate::engine::middleware::service::Context,
+        input: crate::engine::middleware::service::Value,
+        next: crate::engine::middleware::service::Next,
+    ) -> BoxFuture<
+        'static,
+        Result<
+            crate::engine::middleware::service::Value,
+            crate::engine::middleware::service::Error,
+        >,
+    > {
+        self.plan.handle_service(context, input, next)
+    }
+
+    #[must_use]
+    pub fn has_websocket(&self) -> bool {
+        self.plan.has_websocket()
+    }
+    pub fn handle_websocket(
+        &self,
+        context: crate::engine::middleware::websocket::Context,
+        message: crate::engine::middleware::websocket::Message,
+        next: crate::engine::middleware::websocket::Next,
+    ) -> BoxFuture<
+        'static,
+        Result<Option<crate::engine::middleware::websocket::Message>, MiddlewareError>,
+    > {
+        self.plan.handle_websocket(context, message, next)
+    }
+
+    #[must_use]
+    pub fn has_http(&self) -> bool {
+        self.plan.has_http()
+    }
+
+    pub fn handle_http(
+        &self,
+        mut context: crate::engine::middleware::http::Context,
+        request: crate::engine::middleware::http::Request,
+        next: crate::engine::middleware::http::Next,
+    ) -> BoxFuture<'static, Result<crate::engine::middleware::http::Response, MiddlewareError>>
+    {
+        context.plan = Some(self.clone());
+        self.plan.handle_http(context, request, next)
+    }
+
+    #[must_use]
+    pub fn new(plan: Arc<dyn MiddlewarePlan>, generation: ExtensionSetReference) -> Self {
+        Self {
+            plan,
+            _generation: generation,
+        }
+    }
+
+    pub fn handle(
+        &self,
+        mut context: MiddlewareContext,
+        request: MiddlewareRequest,
+        next: MiddlewareNext,
+    ) -> BoxFuture<'static, Result<MiddlewareResponse, MiddlewareError>> {
+        context.plan = Some(self.clone());
+        self.plan.handle(context, request, next)
+    }
+}
+
+impl fmt::Debug for FrozenMiddlewarePlan {
+    fn fmt(&self, formatter: &mut fmt::Formatter<'_>) -> fmt::Result {
+        formatter
+            .debug_struct("FrozenMiddlewarePlan")
+            .finish_non_exhaustive()
+    }
+}

backend/crates/gateway-core/src/engine/middleware/service.rs

@@ -7,7 +7,7 @@ use crate::{
 
 /// 值只在进入插件链时编码；操作的领域校验与副作用由终端所属服务执行
 pub type Value = serde_json::Value;
-pub type Next = super::Next<Value, Value, Error>;
+pub type Next = crate::middleware::Next<Value, Value, Error>;
 
 #[derive(Clone, Debug, PartialEq, Eq, serde::Serialize, serde::Deserialize, thiserror::Error)]
 #[error("{message}")]

backend/crates/gateway-core/src/engine/middleware/websocket.rs

@@ -44,4 +44,4 @@ pub struct Context {
     pub cancellation: CancellationToken,
     pub sender: Arc<dyn Sender>,
 }
-pub type Next = super::Next<Message, Option<Message>, MiddlewareError>;
+pub type Next = crate::middleware::Next<Message, Option<Message>, MiddlewareError>;

backend/crates/gateway-core/src/engine/mod.rs

@@ -101,17 +101,7 @@ pub enum AttemptTrigger {
     AccountRetry,
 }
 
-/// Provider 可为当前 attempt 选择的请求局部传输档位
-#[derive(Debug, Clone, Copy, Default, PartialEq, Eq, Hash)]
-pub enum AttemptTransport {
-    /// 使用 Provider 的默认传输策略
-    #[default]
-    Default,
-    /// 固定账号重试 Provider 的首选传输；序号由 Provider 的独立预算驱动
-    Retry(NonZeroU32),
-    /// 使用 Provider 定义的备用传输
-    Fallback,
-}
+pub use crate::upstream::AttemptTransport;
 
 impl AttemptTrigger {
     #[must_use]
@@ -888,8 +878,8 @@ pub struct ProbeFailure {
     pub latency: Duration,
 }
 
-/// `model_requests` 可用的毫秒级阶段耗时
-#[derive(Debug, Clone, Default, PartialEq, Eq)]
+/// 请求与上游响应的耗时事实，统一以毫秒保存但不提高来源精度
+#[derive(Debug, Clone, Default, PartialEq)]
 pub struct ModelRequestTimings {
     pub transport_decision_wait_ms: Option<u64>,
     pub connect_ms: Option<u64>,
@@ -899,6 +889,13 @@ pub struct ModelRequestTimings {
     pub first_text_ms: Option<u64>,
     pub first_token_ms: Option<u64>,
     pub provider_processing_ms: Option<u64>,
+    pub upstream_response_ms: Option<u64>,
+    pub upstream_api_overhead_ms: Option<f64>,
+    pub upstream_engine_ms: Option<f64>,
+    pub upstream_engine_iapi_ttft_ms: Option<f64>,
+    pub upstream_engine_service_ttft_ms: Option<f64>,
+    pub upstream_engine_iapi_tbt_ms: Option<f64>,
+    pub upstream_engine_service_tbt_ms: Option<f64>,
     pub latency_ms: Option<u64>,
 }
 

backend/crates/gateway-core/src/engine/nested.rs

@@ -65,7 +65,7 @@ pub struct NestedModelExecutionRequest {
 /// Core 在绑定时冻结当前
 /// Key 策略；每个实际模型请求仍独立经过准入、预算、账本与计费
 pub struct BoundModelExecutionBinding {
-    pub settings: Option<crate::settings::RequestSettings>,
+    pub settings: Option<crate::routing::request_settings::RequestSettings>,
     pub client_key_id: ClientApiKeyId,
     pub initiating_plugin_instance_id: String,
     /// 身份随父调用或连接回收；每个模型请求使用独立的执行期限

backend/crates/gateway-core/src/engine/observation.rs

@@ -22,8 +22,8 @@ use crate::{
     metering::{CostEstimate, CostSource, Usage},
     operation::OperationKind,
     policy::ClientApiKeyId,
+    routing::extensions::{ExtensionSetId, ExtensionSetReference},
     routing::{AccountGroupId, ConfigRevision, PublicModelId},
-    runtime::extensions::{ExtensionSetId, ExtensionSetReference},
     upstream::UpstreamSendState,
 };
 
@@ -237,7 +237,7 @@ pub enum RequestObservationOutcome {
 }
 
 /// Core 在业务结果确定后产生的一次最终观察，不包含原始正文或凭据
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct RequestObservation {
     event_id: String,
     request_id: ModelRequestId,
@@ -768,8 +768,6 @@ impl ResponseObservation {
     }
 
     pub(super) fn observe_event(&mut self, event: &GatewayEvent) {
-        let elapsed = self.elapsed_ms();
-        observe_event_timing(&mut self.timings, event, elapsed);
         if let GatewayEvent::Usage(observed) = event {
             self.usage.merge(observed);
         }
@@ -806,6 +804,13 @@ impl ResponseObservation {
         self.timings.first_text_ms = None;
         self.timings.first_token_ms = None;
         self.timings.provider_processing_ms = None;
+        self.timings.upstream_response_ms = None;
+        self.timings.upstream_api_overhead_ms = None;
+        self.timings.upstream_engine_ms = None;
+        self.timings.upstream_engine_iapi_ttft_ms = None;
+        self.timings.upstream_engine_service_ttft_ms = None;
+        self.timings.upstream_engine_iapi_tbt_ms = None;
+        self.timings.upstream_engine_service_tbt_ms = None;
     }
 
     pub(super) fn observe_response(&mut self, observation: &ProviderResponseObservation) {
@@ -834,26 +839,26 @@ impl ResponseObservation {
         if let Some(value) = observed.provider_processing_ms {
             self.timings.provider_processing_ms = Some(value);
         }
-    }
-}
-
-fn observe_event_timing(timings: &mut ModelRequestTimings, event: &GatewayEvent, elapsed_ms: u64) {
-    timings.first_event_ms.get_or_insert(elapsed_ms);
-    match event {
-        GatewayEvent::ReasoningDelta(delta) if !delta.text.is_empty() => {
-            timings.first_reasoning_ms.get_or_insert(elapsed_ms);
-            timings.first_token_ms.get_or_insert(elapsed_ms);
+        if let Some(value) = observed.upstream_response_ms {
+            self.timings.upstream_response_ms = Some(value);
         }
-        GatewayEvent::TextDelta(delta) if !delta.text.is_empty() => {
-            timings.first_text_ms.get_or_insert(elapsed_ms);
-            timings.first_token_ms.get_or_insert(elapsed_ms);
+        if let Some(value) = observed.upstream_api_overhead_ms {
+            self.timings.upstream_api_overhead_ms = Some(value);
         }
-        // `response.output_item.added` 会先投影一个空参数的 tool delta；它只是结构帧，
-        // 不能抢在真实工具参数之前成为首个可消费 token
-        GatewayEvent::ToolCallDelta(delta) if !delta.arguments_delta.is_empty() => {
-            timings.first_token_ms.get_or_insert(elapsed_ms);
+        if let Some(value) = observed.upstream_engine_ms {
+            self.timings.upstream_engine_ms = Some(value);
+        }
+        if let Some(value) = observed.upstream_engine_iapi_ttft_ms {
+            self.timings.upstream_engine_iapi_ttft_ms = Some(value);
+        }
+        if let Some(value) = observed.upstream_engine_service_ttft_ms {
+            self.timings.upstream_engine_service_ttft_ms = Some(value);
+        }
+        if let Some(value) = observed.upstream_engine_iapi_tbt_ms {
+            self.timings.upstream_engine_iapi_tbt_ms = Some(value);
+        }
+        if let Some(value) = observed.upstream_engine_service_tbt_ms {
+            self.timings.upstream_engine_service_tbt_ms = Some(value);
         }
-        GatewayEvent::CalculatedCost(_) | GatewayEvent::ProviderCost(_) => {}
-        _ => {}
     }
 }

backend/crates/gateway-core/src/engine/policy.rs

@@ -19,8 +19,8 @@ use crate::{
     identity::ProviderKind,
     operation::Operation,
     policy::ClientApiKeyId,
+    routing::extensions::{ExtensionSetId, ExtensionSetReference},
     routing::{AccountGroupId, PublicModelId},
-    runtime::extensions::{ExtensionSetId, ExtensionSetReference},
 };
 
 /// 模型路由插件的一次输入；正文仍保存在 `operation` 中，由 Runtime 按权限投影

backend/crates/gateway-core/src/engine/upstream_adapter.rs

@@ -15,7 +15,7 @@ use crate::{
     metering::{CalculatedCost, Usage},
     operation::Operation,
     routing::UpstreamModelId,
-    runtime::extensions::ExtensionSetReference,
+    routing::extensions::ExtensionSetReference,
 };
 
 use super::{

backend/crates/gateway-core/src/error/mod.rs

@@ -144,7 +144,7 @@ pub enum PreDeliveryRetry {
     /// 可靠 NotSent 的建连恢复；次数与时间由 Core 的请求级预算裁决
     SameAccountConnectionRetry {
         /// Provider 明确指定实际失败的传输，避免 HTTP 恢复重新进入 WS
-        transport: crate::engine::AttemptTransport,
+        transport: crate::upstream::AttemptTransport,
     },
     /// 固定本次账号，并要求 Provider 使用备用传输
     SameAccountTransportFallback,
@@ -666,7 +666,7 @@ impl ProviderError {
     }
 
     #[must_use]
-    pub fn with_connection_retry(mut self, transport: crate::engine::AttemptTransport) -> Self {
+    pub fn with_connection_retry(mut self, transport: crate::upstream::AttemptTransport) -> Self {
         self.pre_delivery_retry = Some(Box::new(PreDeliveryRetry::SameAccountConnectionRetry {
             transport,
         }));

backend/crates/gateway-core/src/event.rs

@@ -479,12 +479,14 @@ impl UpstreamHttpVersion {
     }
 }
 
-/// Provider transport 边界测得的时间
+/// Provider transport 观测与上游返回的响应耗时
 ///
 /// `first_*_ms` 是相对 `AttemptContext::timing_started_at()` 的请求级偏移，
-/// 与 Core 总耗时使用同一起点并包含选号与重试等待；其余字段是各阶段独立耗时
+/// 与 Core 总耗时使用同一起点并包含选号与重试等待
+/// `upstream_response_ms` 使用上游的响应计时边界，其余字段是各阶段独立耗时
+/// 首事件与输出阶段时间由 Provider 独占采集，Core 不从 canonical 事件补记
 /// 首字边界由 Provider 协议定义，缺失时保留 `None`，不使用首包代替
-#[derive(Debug, Clone, Copy, Default, PartialEq, Eq)]
+#[derive(Debug, Clone, Copy, Default, PartialEq)]
 pub struct ProviderResponseTimings {
     pub transport_decision_wait_ms: Option<u64>,
     pub connect_ms: Option<u64>,
@@ -494,6 +496,16 @@ pub struct ProviderResponseTimings {
     pub first_text_ms: Option<u64>,
     pub first_token_ms: Option<u64>,
     pub provider_processing_ms: Option<u64>,
+    /// 上游返回的单次响应创建到完成耗时，不使用本地时钟补齐
+    pub upstream_response_ms: Option<u64>,
+    /// 上游排除引擎与客户端工具时间后的 API 耗时
+    pub upstream_api_overhead_ms: Option<f64>,
+    /// 上游引擎服务总耗时；以下专项指标保留毫秒小数及各自内部计量层级
+    pub upstream_engine_ms: Option<f64>,
+    pub upstream_engine_iapi_ttft_ms: Option<f64>,
+    pub upstream_engine_service_ttft_ms: Option<f64>,
+    pub upstream_engine_iapi_tbt_ms: Option<f64>,
+    pub upstream_engine_service_tbt_ms: Option<f64>,
 }
 
 /// Provider 已筛选的响应观测 JSON
@@ -572,7 +584,7 @@ impl fmt::Debug for ProviderResponseHeader {
 }
 
 /// Core 消费的 Provider 执行观测，不会原样进入客户端响应
-#[derive(Debug, Clone, PartialEq, Eq)]
+#[derive(Debug, Clone, PartialEq)]
 pub struct ProviderResponseObservation {
     transport: UpstreamTransport,
     http_version: Option<UpstreamHttpVersion>,

backend/crates/gateway-core/src/lib.rs

@@ -128,7 +128,7 @@ pub struct CoreStartup {
     providers: ProviderRegistry,
     request_observers: Option<engine::observation::RequestObserverExtensionIndex>,
     request_policies: Option<engine::policy::RequestPolicyExtensionIndex>,
-    middlewares: Option<engine::middleware::MiddlewareExtensionIndex>,
+    execution_extensions: Option<engine::extensions::ExecutionExtensionIndex>,
     frontend_authentication: Option<engine::authentication::FrontendAuthenticationExtensionIndex>,
     control: CoreControlPlaneStartup,
 }
@@ -164,7 +164,7 @@ impl CoreStartup {
             self.providers,
             self.request_observers,
             self.request_policies,
-            self.middlewares,
+            self.execution_extensions,
             self.frontend_authentication,
         ));
         let execution: Arc<dyn ExecutionService> = service.clone();
@@ -194,7 +194,7 @@ fn build_execution_service(
     providers: ProviderRegistry,
     request_observers: Option<engine::observation::RequestObserverExtensionIndex>,
     request_policies: Option<engine::policy::RequestPolicyExtensionIndex>,
-    middlewares: Option<engine::middleware::MiddlewareExtensionIndex>,
+    execution_extensions: Option<engine::extensions::ExecutionExtensionIndex>,
     frontend_authentication: Option<engine::authentication::FrontendAuthenticationExtensionIndex>,
 ) -> DefaultExecutionService {
     let mut service = DefaultExecutionService::new(
@@ -215,8 +215,8 @@ fn build_execution_service(
     if let Some(request_policies) = request_policies {
         service = service.with_request_policies(request_policies);
     }
-    if let Some(middlewares) = middlewares {
-        service = service.with_middlewares(middlewares);
+    if let Some(execution_extensions) = execution_extensions {
+        service = service.with_execution_extensions(execution_extensions);
     }
     if let Some(frontend_authentication) = frontend_authentication {
         service = service.with_frontend_authentication(frontend_authentication);
@@ -276,10 +276,10 @@ impl CoreBundle {
 pub async fn initialize(
     ports: CoreStorePorts,
     providers: ProviderRegistry,
-    extensions: Option<Arc<dyn runtime::extensions::ExtensionPreparationPort>>,
+    extensions: Option<Arc<dyn routing::extensions::ExtensionPreparationPort>>,
     request_observers: Option<engine::observation::RequestObserverExtensionIndex>,
     request_policies: Option<engine::policy::RequestPolicyExtensionIndex>,
-    middlewares: Option<engine::middleware::MiddlewareExtensionIndex>,
+    execution_extensions: Option<engine::extensions::ExecutionExtensionIndex>,
     frontend_authentication: Option<engine::authentication::FrontendAuthenticationExtensionIndex>,
 ) -> Result<CoreBundle, CoreError> {
     prepare(
@@ -288,7 +288,7 @@ pub async fn initialize(
         extensions,
         request_observers,
         request_policies,
-        middlewares,
+        execution_extensions,
         frontend_authentication,
     )
     .activate()
@@ -300,10 +300,10 @@ pub async fn initialize(
 pub fn prepare(
     ports: CoreStorePorts,
     providers: ProviderRegistry,
-    extensions: Option<Arc<dyn runtime::extensions::ExtensionPreparationPort>>,
+    extensions: Option<Arc<dyn routing::extensions::ExtensionPreparationPort>>,
     request_observers: Option<engine::observation::RequestObserverExtensionIndex>,
     request_policies: Option<engine::policy::RequestPolicyExtensionIndex>,
-    middlewares: Option<engine::middleware::MiddlewareExtensionIndex>,
+    execution_extensions: Option<engine::extensions::ExecutionExtensionIndex>,
     frontend_authentication: Option<engine::authentication::FrontendAuthenticationExtensionIndex>,
 ) -> CoreStartup {
     let control = prepare_control_plane(ports.clone(), providers.clone(), extensions);
@@ -312,7 +312,7 @@ pub fn prepare(
         providers,
         request_observers,
         request_policies,
-        middlewares,
+        execution_extensions,
         frontend_authentication,
         control,
     }
@@ -377,7 +377,7 @@ pub struct CoreCommandPlaneStartup {
     providers: ProviderRegistry,
     request_observers: Option<engine::observation::RequestObserverExtensionIndex>,
     request_policies: Option<engine::policy::RequestPolicyExtensionIndex>,
-    middlewares: Option<engine::middleware::MiddlewareExtensionIndex>,
+    execution_extensions: Option<engine::extensions::ExecutionExtensionIndex>,
     control: CoreControlPlaneStartup,
 }
 
@@ -407,7 +407,7 @@ impl CoreCommandPlaneStartup {
                 self.providers,
                 self.request_observers,
                 self.request_policies,
-                self.middlewares,
+                self.execution_extensions,
                 None,
             )
             .with_snapshot_refresh(Arc::clone(&control.publisher)),
@@ -449,10 +449,10 @@ impl CoreCommandPlaneBundle {
 pub fn prepare_command_plane(
     ports: CoreStorePorts,
     providers: ProviderRegistry,
-    extensions: Option<Arc<dyn runtime::extensions::ExtensionPreparationPort>>,
+    extensions: Option<Arc<dyn routing::extensions::ExtensionPreparationPort>>,
     request_observers: Option<engine::observation::RequestObserverExtensionIndex>,
     request_policies: Option<engine::policy::RequestPolicyExtensionIndex>,
-    middlewares: Option<engine::middleware::MiddlewareExtensionIndex>,
+    execution_extensions: Option<engine::extensions::ExecutionExtensionIndex>,
 ) -> CoreCommandPlaneStartup {
     let control = prepare_control_plane(ports.clone(), providers.clone(), extensions);
     CoreCommandPlaneStartup {
@@ -460,27 +460,17 @@ pub fn prepare_command_plane(
         providers,
         request_observers,
         request_policies,
-        middlewares,
+        execution_extensions,
         control,
     }
 }
 
-pub async fn initialize_control_plane(
-    ports: CoreStorePorts,
-    providers: ProviderRegistry,
-    extensions: Option<Arc<dyn runtime::extensions::ExtensionPreparationPort>>,
-) -> Result<CoreControlPlaneBundle, CoreError> {
-    prepare_control_plane(ports, providers, extensions)
-        .activate()
-        .await
-}
-
 /// 仅构造管理命令所需快照能力，不恢复准入、不发现目录且不启动 Worker
 #[must_use]
 pub fn prepare_control_plane(
     ports: CoreStorePorts,
     providers: ProviderRegistry,
-    extensions: Option<Arc<dyn runtime::extensions::ExtensionPreparationPort>>,
+    extensions: Option<Arc<dyn routing::extensions::ExtensionPreparationPort>>,
 ) -> CoreControlPlaneStartup {
     let mut compiler = RuntimeSnapshotCompiler::new(ports.snapshots, Arc::new(providers));
     if let Some(extensions) = extensions {