feat: refine admin controls and overload protection

Refine settings navigation, selective pricing sync, model deletion, and admin interactions. Consolidate unreleased migrations and keep Fast policy scoped to groups. Increase confirmed capacity-rejection feedback without changing ordinary failure recovery.

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

README.md

@@ -86,9 +86,6 @@ API Key 持有者可在同一登录页切换登录身份，进入 `/key-usage`
 2. 按需建立账号分组，再创建客户端密钥并选择可用分组。**不选分组表示可使用全部账号**。
 3. 打开密钥的「使用密钥」，复制客户端配置。
 
-账号自动冻结默认关闭，可在「系统设置 → 运行参数」中启用。启用后，账号频繁命中上游容量错误时
-会暂停调度并显示为限流中；若开启恢复前探测，冷却到期后需探测成功才恢复。
-
 ## 客户端接入
 
 **Codex CLI / 桌面端**：在「使用密钥」中按操作系统复制配置，或通过 CCSwitch 导入。

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

@@ -1,7 +1,7 @@
 //! 全局模型定价的管理合同。
 
 use gateway_core::metering::{ModelPriceOverride, PricingOverrides};
-use std::collections::BTreeMap;
+use std::collections::{BTreeMap, BTreeSet};
 
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct PricingCatalog {
@@ -25,11 +25,21 @@ pub struct PricingSyncPreview {
     pub skipped: Vec<String>,
 }
 
+#[derive(Debug, Clone, PartialEq, Eq, serde::Deserialize)]
+#[serde(rename_all = "camelCase", deny_unknown_fields)]
+pub struct SyncPricing {
+    pub preview: PricingSyncPreview,
+    pub models: BTreeMap<String, BTreeSet<String>>,
+}
+
+pub type PricingSyncChanges = BTreeMap<String, BTreeMap<String, Option<ModelPriceOverride>>>;
+
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub enum PricingChange {
     Replace(ModelPriceOverride),
     Multiplier(u32),
     Reset,
+    Delete,
 }
 
 #[derive(Debug, Clone, PartialEq, Eq)]

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

@@ -18,7 +18,6 @@ pub use gateway_core::account::RotationStrategy;
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct RuntimeSettings {
     pub openai_client_profile: Option<gateway_core::account::OpaqueProviderData>,
-    pub disable_fast: bool,
     pub config_revision: Revision,
     pub request_location_enabled: bool,
     pub request_location: gateway_core::account::RequestLocation,
@@ -51,7 +50,6 @@ pub struct RuntimeSettings {
 #[derive(Debug, Clone, PartialEq, Eq)]
 pub struct ReplaceRuntimeSettings {
     pub openai_client_profile: Option<gateway_core::account::OpaqueProviderData>,
-    pub disable_fast: Option<bool>,
     pub request_location_enabled: bool,
     pub request_location: gateway_core::account::RequestLocation,
     pub model_mappings: ModelMappings,

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

@@ -420,7 +420,7 @@ pub trait SettingsStore: Send + Sync {
     async fn load_pricing(&self) -> AdminStoreResult<crate::model::pricing::StoredPricing>;
     async fn sync_pricing(
         &self,
-        prices: gateway_core::metering::PricingOverrides,
+        changes: crate::model::pricing::PricingSyncChanges,
         context: &MutationContext,
     ) -> AdminStoreResult<crate::model::Revision>;
     async fn update_pricing(

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

@@ -29,7 +29,7 @@ pub trait SettingsService: Send + Sync {
     async fn sync_pricing(
         &self,
         context: &MutationContext,
-        preview: crate::model::pricing::PricingSyncPreview,
+        command: crate::model::pricing::SyncPricing,
     ) -> Result<(), AdminError>;
     async fn pricing(&self) -> Result<crate::model::pricing::PricingCatalog, AdminError>;
     async fn update_pricing(
@@ -103,17 +103,55 @@ impl SettingsService for DefaultSettingsService {
     async fn sync_pricing(
         &self,
         context: &MutationContext,
-        preview: crate::model::pricing::PricingSyncPreview,
+        command: crate::model::pricing::SyncPricing,
     ) -> Result<(), AdminError> {
-        let current = self.pricing_source.fetch().await?;
-        if current != preview {
+        let count = command
+            .models
+            .values()
+            .map(std::collections::BTreeSet::len)
+            .sum::<usize>();
+        if count == 0
+            || count > 10_000
+            || command
+                .models
+                .values()
+                .any(std::collections::BTreeSet::is_empty)
+        {
+            return Err(AdminError::invalid("请选择 1 至 10000 个模型"));
+        }
+        let mut current = self.pricing_source.fetch().await?;
+        if current != command.preview {
             return Err(AdminError::invalid(
                 "models.dev 价目已变化，请重新预览后确认",
             ));
         }
+        let stored = self
+            .store
+            .load_pricing()
+            .await
+            .map_err(|error| map_store_error(error, "model pricing"))?;
+        let mut changes = crate::model::pricing::PricingSyncChanges::new();
+        for (provider, models) in command.models {
+            let selected = changes.entry(provider.clone()).or_default();
+            for model in models {
+                let price = current
+                    .prices
+                    .get_mut(&provider)
+                    .and_then(|prices| prices.remove(&model));
+                if price.is_none()
+                    && !stored
+                        .synced
+                        .get(&provider)
+                        .is_some_and(|prices| prices.contains_key(&model))
+                {
+                    return Err(AdminError::invalid("所选模型不在来源价目中，请重新预览"));
+                }
+                selected.insert(model, price);
+            }
+        }
         let revision = self
             .store
-            .sync_pricing(current.prices, context)
+            .sync_pricing(changes, context)
             .await
             .map_err(|error| map_store_error(error, "model pricing sync"))?;
         publish_committed(self.snapshot.as_ref(), revision).await
@@ -153,6 +191,16 @@ impl SettingsService for DefaultSettingsService {
         {
             return Err(AdminError::invalid("Provider、模型 ID 或批量数量不合法"));
         }
+        if command.change == PricingChange::Delete
+            && command.models.iter().any(|model| {
+                catalog
+                    .defaults
+                    .get(&command.provider)
+                    .is_some_and(|models| models.contains_key(model))
+            })
+        {
+            return Err(AdminError::invalid("内置价目不能删除，仅支持人工覆盖"));
+        }
         let defaults = gateway_core::metering::merge_pricing(catalog.defaults, &catalog.synced);
         let defaults = defaults.get(&command.provider);
         match &command.change {

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

@@ -976,7 +976,7 @@ impl SettingsStore for StaticSettingsStore {
     }
     async fn sync_pricing(
         &self,
-        _: gateway_core::metering::PricingOverrides,
+        _: gateway_admin::model::pricing::PricingSyncChanges,
         _: &MutationContext,
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         panic!("unexpected pricing sync")
@@ -991,7 +991,6 @@ impl SettingsStore for StaticSettingsStore {
     async fn load_runtime_settings(&self) -> AdminStoreResult<RuntimeSettings> {
         Ok(RuntimeSettings {
             openai_client_profile: None,
-            disable_fast: false,
             request_location_enabled: false,
             request_location: Default::default(),
             config_revision: revision(1),

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

@@ -26,7 +26,6 @@ use super::accounts::{FakeAccountStore, FakeProviderAdmin, account_record, event
 fn runtime_settings(enabled: bool, probe_enabled: bool, adaptive: bool) -> RuntimeSettings {
     RuntimeSettings {
         openai_client_profile: None,
-        disable_fast: false,
         request_location_enabled: false,
         request_location: Default::default(),
         config_revision: revision(1),
@@ -67,7 +66,7 @@ impl SettingsStore for FreezeSettingsStore {
     }
     async fn sync_pricing(
         &self,
-        _: gateway_core::metering::PricingOverrides,
+        _: gateway_admin::model::pricing::PricingSyncChanges,
         _: &MutationContext,
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         panic!("unexpected pricing sync")

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

@@ -711,7 +711,7 @@ impl SettingsStore for UnavailableStore {
     }
     async fn sync_pricing(
         &self,
-        _: gateway_core::metering::PricingOverrides,
+        _: gateway_admin::model::pricing::PricingSyncChanges,
         _: &MutationContext,
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         panic!("unexpected pricing sync")

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

@@ -758,7 +758,7 @@ impl SettingsStore for FixtureSettingsStore {
     }
     async fn sync_pricing(
         &self,
-        _: gateway_core::metering::PricingOverrides,
+        _: gateway_admin::model::pricing::PricingSyncChanges,
         _: &MutationContext,
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         panic!("unexpected pricing sync")
@@ -773,7 +773,6 @@ impl SettingsStore for FixtureSettingsStore {
     async fn load_runtime_settings(&self) -> AdminStoreResult<RuntimeSettings> {
         Ok(RuntimeSettings {
             openai_client_profile: None,
-            disable_fast: false,
             request_location_enabled: false,
             request_location: Default::default(),
             config_revision: Revision::new(1).expect("revision"),

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

@@ -20,7 +20,7 @@ impl SettingsStore for UnusedSettingsStore {
     }
     async fn sync_pricing(
         &self,
-        _: gateway_core::metering::PricingOverrides,
+        _: gateway_admin::model::pricing::PricingSyncChanges,
         _: &MutationContext,
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         panic!("unexpected pricing sync")
@@ -79,7 +79,6 @@ async fn settings_should_reject_zero_refresh_margin_before_store_call() {
             },
             ReplaceRuntimeSettings {
                 openai_client_profile: None,
-                disable_fast: None,
                 request_location_enabled: false,
                 request_location: Default::default(),
                 model_mappings: Default::default(),

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

@@ -35,7 +35,6 @@ pub type ModelMappings = BTreeMap<String, String>;
 #[serde(rename_all = "camelCase")]
 pub struct RuntimeSettingsView {
     pub openai_client_profile: Option<serde_json::Map<String, serde_json::Value>>,
-    pub disable_fast: bool,
     pub request_location_enabled: bool,
     pub request_location: gateway_core::account::RequestLocation,
     pub model_mappings: ModelMappings,
@@ -69,7 +68,6 @@ pub struct RuntimeSettingsView {
 pub struct UpdateRuntimeSettingsRequest {
     #[serde(default, deserialize_with = "deserialize_profile_update")]
     pub openai_client_profile: Option<serde_json::Map<String, serde_json::Value>>,
-    pub disable_fast: Option<bool>,
     pub request_location_enabled: bool,
     pub request_location: gateway_core::account::RequestLocation,
     pub model_mappings: ModelMappings,
@@ -183,7 +181,6 @@ impl UpdateRuntimeSettingsRequest {
             openai_client_profile: self
                 .openai_client_profile
                 .map(gateway_core::account::OpaqueProviderData::new),
-            disable_fast: self.disable_fast,
             request_location_enabled: self.request_location_enabled,
             request_location: self
                 .request_location
@@ -228,7 +225,6 @@ impl From<RuntimeSettings> for RuntimeSettingsView {
             openai_client_profile: settings
                 .openai_client_profile
                 .map(gateway_core::account::OpaqueProviderData::into_inner),
-            disable_fast: settings.disable_fast,
             request_location_enabled: settings.request_location_enabled,
             request_location: settings.request_location,
             model_mappings: wire_model_mappings(settings.model_mappings),
@@ -429,6 +425,7 @@ enum PricingChangeRequest {
         multiplier_bps: u32,
     },
     Reset {},
+    Delete {},
 }
 
 async fn pricing<S>(
@@ -475,7 +472,7 @@ where
 async fn sync_pricing<S>(
     auth: AdminAuth,
     State(state): State<S>,
-    AdminJson(preview): AdminJson<gateway_admin::model::pricing::PricingSyncPreview>,
+    AdminJson(command): AdminJson<gateway_admin::model::pricing::SyncPricing>,
 ) -> Result<impl IntoResponse, AdminError>
 where
     S: SessionState + Send + Sync,
@@ -483,7 +480,7 @@ where
     state
         .admin_services()
         .settings()
-        .sync_pricing(&auth.context().mutation_context(), preview)
+        .sync_pricing(&auth.context().mutation_context(), command)
         .await
         .map_err(map_service_error)?;
     Ok(AdminResponse::new(
@@ -507,6 +504,7 @@ where
             PricingChange::Multiplier(multiplier_bps)
         }
         PricingChangeRequest::Reset {} => PricingChange::Reset,
+        PricingChangeRequest::Delete {} => PricingChange::Delete,
     };
     state
         .admin_services()

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

@@ -446,11 +446,21 @@ impl SettingsStore for MemorySettingsStore {
     }
     async fn sync_pricing(
         &self,
-        prices: gateway_core::metering::PricingOverrides,
+        changes: gateway_admin::model::pricing::PricingSyncChanges,
         _: &MutationContext,
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         let mut pricing = self.pricing.lock().expect("pricing");
-        pricing.synced = prices;
+        for (provider, models) in changes {
+            let stored = pricing.synced.entry(provider).or_default();
+            for (model, price) in models {
+                if let Some(price) = price {
+                    stored.insert(model, price);
+                } else {
+                    stored.remove(&model);
+                }
+            }
+        }
+        pricing.synced.retain(|_, models| !models.is_empty());
         pricing.synced_at = Some(Utc::now());
         let mut settings = self.settings.lock().expect("settings");
         settings.config_revision = next_revision(settings.config_revision);
@@ -463,12 +473,20 @@ impl SettingsStore for MemorySettingsStore {
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         use gateway_admin::model::pricing::PricingChange;
         let mut pricing = self.pricing.lock().expect("pricing");
-        let models = pricing.overrides.entry(command.provider).or_default();
+        let gateway_admin::model::pricing::StoredPricing {
+            overrides, synced, ..
+        } = &mut *pricing;
+        let models = overrides.entry(command.provider.clone()).or_default();
+        let source = synced.entry(command.provider).or_default();
         for model in command.models {
             match &command.change {
                 PricingChange::Reset => {
                     models.remove(&model);
                 }
+                PricingChange::Delete => {
+                    models.remove(&model);
+                    source.remove(&model);
+                }
                 PricingChange::Replace(p) => {
                     models.insert(model, p.clone());
                 }
@@ -503,7 +521,6 @@ impl SettingsStore for MemorySettingsStore {
         let mut settings = self.settings.lock().expect("settings");
         let updated = RuntimeSettings {
             openai_client_profile: None,
-            disable_fast: command.disable_fast.unwrap_or(settings.disable_fast),
             request_location_enabled: command.request_location_enabled,
             request_location: command.request_location,
             config_revision: next_revision(settings.config_revision),
@@ -1243,6 +1260,15 @@ impl UnusedProvider {
 
 #[async_trait]
 impl ProviderAdmin for UnusedProvider {
+    fn pricing_catalog(&self) -> gateway_admin::model::pricing::ProviderPricingCatalog {
+        serde_json::from_value(serde_json::json!({
+            "builtin-model": {
+                "multiplierBps": 10000,
+                "bands": {"standard": {"input": "1", "output": "2", "cacheRead": "0", "cacheWrite": "1"}}
+            }
+        })).unwrap()
+    }
+
     fn provider_kind(&self) -> &ProviderKind {
         &self.kind
     }
@@ -1409,7 +1435,6 @@ fn test_runtime_settings() -> RuntimeSettings {
     ]);
     RuntimeSettings {
         openai_client_profile: None,
-        disable_fast: false,
         request_location_enabled: false,
         request_location: Default::default(),
         config_revision: Revision::new(7).expect("revision"),

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

@@ -37,7 +37,6 @@ async fn response_json(response: axum::response::Response) -> Value {
 
 fn update_body() -> Value {
     json!({
-        "disableFast": false,
         "requestLocationEnabled": false,
         "requestLocation": {"country":"US", "region":"Ohio", "city":"Piketon", "timezone":"America/New_York"},
         "modelMappings": {
@@ -104,7 +103,6 @@ fn settings_response_should_cover_the_full_runtime_settings_contract() {
 
     let settings = RuntimeSettings {
         openai_client_profile: None,
-        disable_fast: false,
         request_location_enabled: false,
         request_location: Default::default(),
         config_revision: Revision::new(7).expect("revision"),
@@ -150,7 +148,6 @@ fn settings_response_should_cover_the_full_runtime_settings_contract() {
         value,
         json!({
             "openaiClientProfile": null,
-            "disableFast": false,
         "requestLocationEnabled": false,
         "requestLocation": {"country":"US", "region":"Ohio", "city":"Piketon", "timezone":"America/New_York"},
             "modelMappings": {
@@ -205,7 +202,6 @@ fn settings_request_and_response_fields_should_stay_in_lockstep() {
         .collect();
     let settings = RuntimeSettings {
         openai_client_profile: None,
-        disable_fast: false,
         request_location_enabled: false,
         request_location: Default::default(),
         config_revision: Revision::new(7).expect("revision"),
@@ -610,17 +606,14 @@ fn decompression_setting_should_reject_invalid_values() {
 }
 
 #[tokio::test]
-async fn disable_fast_settings_updates_preserve_omitted_values() {
+async fn settings_reject_removed_global_fast_policy() {
     let fixture = AdminTestFixture::new().await;
     fixture.auth.insert_session("valid-session");
     let app = app(fixture.state());
-    for (value, expected) in [(Some(true), true), (None, true), (Some(false), false)] {
+    let initial = fixture.settings.settings.lock().unwrap().clone();
+    for value in [json!(true), json!(false), Value::Null] {
         let mut body = update_body();
-        if let Some(value) = value {
-            body["disableFast"] = json!(value);
-        } else {
-            body.as_object_mut().unwrap().remove("disableFast");
-        }
+        body["disableFast"] = value;
         let response = app
             .clone()
             .oneshot(request(
@@ -630,17 +623,19 @@ async fn disable_fast_settings_updates_preserve_omitted_values() {
             ))
             .await
             .unwrap();
-        assert_eq!(response.status(), StatusCode::OK);
-        let response = app
-            .clone()
-            .oneshot(request(Method::GET, "/api/admin/settings", None))
-            .await
-            .unwrap();
-        assert_eq!(
-            response_json(response).await["data"]["disableFast"],
-            expected
-        );
+        assert_eq!(response.status(), StatusCode::UNPROCESSABLE_ENTITY);
+        assert_eq!(*fixture.settings.settings.lock().unwrap(), initial);
     }
+    let response = app
+        .oneshot(request(Method::GET, "/api/admin/settings", None))
+        .await
+        .unwrap();
+    assert_eq!(response.status(), StatusCode::OK);
+    assert!(
+        response_json(response).await["data"]
+            .get("disableFast")
+            .is_none()
+    );
 }
 
 #[test]
@@ -685,7 +680,7 @@ async fn pricing_routes_keep_manual_overrides_during_sync_and_reset_to_source()
         .oneshot(request(
             Method::POST,
             "/api/admin/settings/pricing/sync",
-            Some(approved),
+            Some(json!({"preview": approved, "models": {"openai": ["gpt-5.4"]}})),
         ))
         .await
         .unwrap();
@@ -745,6 +740,88 @@ async fn pricing_routes_keep_manual_overrides_during_sync_and_reset_to_source()
     assert!(result["data"]["synced"]["openai"]["gpt-5.4"].is_object());
 }
 
+#[tokio::test]
+async fn pricing_delete_rejects_builtin_models_even_with_overrides_without_partial_writes() {
+    for overridden in [false, true] {
+        let fixture = AdminTestFixture::new().await;
+        fixture.auth.insert_session("valid-session");
+        let app = app(fixture.state());
+        {
+            let mut stored = fixture.settings.pricing.lock().unwrap();
+            stored.synced = serde_json::from_value(json!({
+                "openai": {"builtin-model": custom_pricing(), "custom-model": custom_pricing()}
+            }))
+            .unwrap();
+            if overridden {
+                stored.overrides = stored.synced.clone();
+            }
+        }
+        let before = fixture.settings.pricing.lock().unwrap().clone();
+        let revision = fixture.settings.settings.lock().unwrap().config_revision;
+        for models in [
+            json!(["builtin-model"]),
+            json!(["custom-model", "builtin-model"]),
+        ] {
+            let response = app
+                .clone()
+                .oneshot(request(
+                    Method::POST,
+                    "/api/admin/settings/pricing/update",
+                    Some(
+                        json!({"provider":"openai", "models":models, "change":{"action":"delete"}}),
+                    ),
+                ))
+                .await
+                .unwrap();
+            assert_eq!(response.status(), StatusCode::BAD_REQUEST);
+            assert_eq!(*fixture.settings.pricing.lock().unwrap(), before);
+            assert_eq!(
+                fixture.settings.settings.lock().unwrap().config_revision,
+                revision
+            );
+        }
+    }
+}
+
+#[tokio::test]
+async fn pricing_delete_removes_selected_nonbuiltin_models_from_both_layers() {
+    let fixture = AdminTestFixture::new().await;
+    fixture.auth.insert_session("valid-session");
+    let app = app(fixture.state());
+    {
+        let mut stored = fixture.settings.pricing.lock().unwrap();
+        stored.synced = serde_json::from_value(json!({
+            "openai": {"shared":custom_pricing(), "synced-only":custom_pricing(), "untouched":custom_pricing()},
+            "xai": {"shared":custom_pricing()}
+        })).unwrap();
+        stored.overrides = serde_json::from_value(json!({
+            "openai": {"shared":custom_pricing(), "custom-only":custom_pricing(), "untouched":custom_pricing()},
+            "xai": {"shared":custom_pricing()}
+        })).unwrap();
+    }
+    for _ in 0..2 {
+        let response = app.clone().oneshot(request(
+            Method::POST,
+            "/api/admin/settings/pricing/update",
+            Some(json!({
+                "provider":"openai", "models":["shared", "synced-only", "custom-only"], "change":{"action":"delete"}
+            })),
+        )).await.unwrap();
+        assert_eq!(response.status(), StatusCode::OK);
+    }
+    let result = response_json(
+        app.oneshot(request(Method::GET, "/api/admin/settings/pricing", None))
+            .await
+            .unwrap(),
+    )
+    .await;
+    let expected =
+        json!({"openai":{"untouched":custom_pricing()},"xai":{"shared":custom_pricing()}});
+    assert_eq!(result["data"]["overrides"], expected);
+    assert_eq!(result["data"]["synced"], expected);
+    assert!(result["data"]["defaults"]["openai"]["builtin-model"].is_object());
+}
+
 #[tokio::test]
 async fn pricing_rejects_invalid_edits_and_tampered_sync_without_writes() {
     let fixture = AdminTestFixture::new().await;
@@ -787,7 +864,7 @@ async fn pricing_rejects_invalid_edits_and_tampered_sync_without_writes() {
         .oneshot(request(
             Method::POST,
             "/api/admin/settings/pricing/sync",
-            Some(json!({"prices":{},"skipped":[]})),
+            Some(json!({"preview":{"prices":{},"skipped":[]},"models":{"openai":["gpt-5.4"]}})),
         ))
         .await
         .unwrap();
@@ -802,6 +879,138 @@ async fn pricing_rejects_invalid_edits_and_tampered_sync_without_writes() {
     assert_eq!(result["data"]["synced"], json!({}));
 }
 
+#[tokio::test]
+async fn pricing_sync_only_updates_selected_provider_models() {
+    let fixture = AdminTestFixture::new().await;
+    fixture.auth.insert_session("valid-session");
+    let old = custom_pricing();
+    {
+        let mut pricing = fixture.settings.pricing.lock().unwrap();
+        pricing.synced = serde_json::from_value(json!({
+            "openai": {"gpt-5.4": old, "untouched": old},
+            "xai": {"gpt-5.4": old}
+        }))
+        .unwrap();
+        pricing.overrides = serde_json::from_value(json!({"openai": {"gpt-5.4": old}})).unwrap();
+    }
+    let app = app(fixture.state());
+    let preview = response_json(
+        app.clone()
+            .oneshot(request(
+                Method::POST,
+                "/api/admin/settings/pricing/sync/preview",
+                None,
+            ))
+            .await
+            .unwrap(),
+    )
+    .await;
+    let response = app
+        .oneshot(request(
+            Method::POST,
+            "/api/admin/settings/pricing/sync",
+            Some(json!({"preview": preview["data"], "models": {"openai": ["gpt-5.4"]}})),
+        ))
+        .await
+        .unwrap();
+    assert_eq!(response.status(), StatusCode::OK);
+    let pricing = fixture.settings.pricing.lock().unwrap();
+    assert_eq!(
+        serde_json::to_value(&pricing.synced).unwrap(),
+        json!({
+            "openai": {"gpt-5.4": preview["data"]["prices"]["openai"]["gpt-5.4"], "untouched": old},
+            "xai": {"gpt-5.4": old}
+        })
+    );
+    assert_eq!(
+        serde_json::to_value(&pricing.overrides).unwrap(),
+        json!({"openai": {"gpt-5.4": old}})
+    );
+}
+
+#[tokio::test]
+async fn pricing_sync_removes_only_selected_retired_source_prices() {
+    let fixture = AdminTestFixture::new().await;
+    fixture.auth.insert_session("valid-session");
+    {
+        let mut pricing = fixture.settings.pricing.lock().unwrap();
+        pricing.synced = serde_json::from_value(json!({
+            "openai": {"retired": custom_pricing(), "untouched": custom_pricing()}
+        }))
+        .unwrap();
+    }
+    let app = app(fixture.state());
+    let preview = response_json(
+        app.clone()
+            .oneshot(request(
+                Method::POST,
+                "/api/admin/settings/pricing/sync/preview",
+                None,
+            ))
+            .await
+            .unwrap(),
+    )
+    .await;
+    let response = app
+        .oneshot(request(
+            Method::POST,
+            "/api/admin/settings/pricing/sync",
+            Some(json!({"preview": preview["data"], "models": {"openai": ["retired"]}})),
+        ))
+        .await
+        .unwrap();
+    assert_eq!(response.status(), StatusCode::OK);
+    let pricing = fixture.settings.pricing.lock().unwrap();
+    assert_eq!(
+        serde_json::to_value(&pricing.synced).unwrap(),
+        json!({
+            "openai": {"untouched": custom_pricing()}
+        })
+    );
+}
+
+#[tokio::test]
+async fn pricing_sync_rejects_empty_unknown_or_oversized_selections_without_writes() {
+    let fixture = AdminTestFixture::new().await;
+    fixture.auth.insert_session("valid-session");
+    let app = app(fixture.state());
+    let preview = response_json(
+        app.clone()
+            .oneshot(request(
+                Method::POST,
+                "/api/admin/settings/pricing/sync/preview",
+                None,
+            ))
+            .await
+            .unwrap(),
+    )
+    .await;
+    for models in [
+        json!({}),
+        json!({"openai": []}),
+        json!({"unknown": ["gpt-5.4"]}),
+        json!({"openai": ["gpt-5.4", "missing"]}),
+        json!({"openai": ["gpt-5.4"], "xai": []}),
+        json!({"openai": (0..10_001).map(|index| format!("model-{index}")).collect::<Vec<_>>()}),
+    ] {
+        let response = app
+            .clone()
+            .oneshot(request(
+                Method::POST,
+                "/api/admin/settings/pricing/sync",
+                Some(json!({"preview": preview["data"], "models": models})),
+            ))
+            .await
+            .unwrap();
+        assert_eq!(response.status(), StatusCode::BAD_REQUEST);
+    }
+    let pricing = fixture.settings.pricing.lock().unwrap();
+    assert_eq!(
+        *pricing,
+        gateway_admin::model::pricing::StoredPricing::default()
+    );
+}
+
 #[tokio::test]
 async fn pricing_endpoints_require_administrator_authentication() {
     let fixture = AdminTestFixture::new().await;
@@ -821,7 +1030,7 @@ async fn pricing_endpoints_require_administrator_authentication() {
         (
             Method::POST,
             "/api/admin/settings/pricing/sync",
-            Some(json!({"prices":{},"skipped":[]})),
+            Some(json!({"preview":{"prices":{},"skipped":[]},"models":{"openai":["gpt-5.4"]}})),
         ),
     ] {
         assert_eq!(

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

@@ -112,6 +112,7 @@ pub struct AccountRuntimeSignals {
 }
 
 const ACCOUNT_FEEDBACK_EWMA_ALPHA: f64 = 0.2;
+const ACCOUNT_CAPACITY_FAILURE_EWMA_ALPHA: f64 = 0.4;
 const ACCOUNT_FAILURE_RATE_HALF_LIFE: Duration = Duration::from_secs(15 * 60);
 const EMPTY_FEEDBACK_SAMPLE: u64 = f64::NAN.to_bits();
 
@@ -120,6 +121,7 @@ const EMPTY_FEEDBACK_SAMPLE: u64 = f64::NAN.to_bits();
 pub enum AccountAttemptFeedback {
     Succeeded { first_output_ms: Option<u64> },
     Failed { first_output_ms: Option<u64> },
+    CapacityRejected { first_output_ms: Option<u64> },
 }
 
 #[derive(Debug, Clone, PartialEq, Eq, Hash)]
@@ -150,13 +152,12 @@ impl DecayingAccountFailureRate {
             * 0.5_f64.powf(elapsed.as_secs_f64() / ACCOUNT_FAILURE_RATE_HALF_LIFE.as_secs_f64())
     }
 
-    fn report_at(&mut self, sample: f64, now: Instant) {
+    fn report_at(&mut self, sample: f64, alpha: f64, now: Instant) {
         let now = self
             .updated_at
             .map_or(now, |updated_at| updated_at.max(now));
         let decayed = self.value_at(now);
-        self.value =
-            ACCOUNT_FEEDBACK_EWMA_ALPHA * sample + (1.0 - ACCOUNT_FEEDBACK_EWMA_ALPHA) * decayed;
+        self.value = alpha * sample + (1.0 - alpha) * decayed;
         self.updated_at = Some(now);
     }
 }
@@ -173,14 +174,21 @@ impl Default for AccountFeedback {
 
 impl AccountFeedback {
     fn report_at(&self, feedback: AccountAttemptFeedback, now: Instant) {
-        let (failure, first_output_ms) = match feedback {
-            AccountAttemptFeedback::Succeeded { first_output_ms } => (0.0, first_output_ms),
-            AccountAttemptFeedback::Failed { first_output_ms } => (1.0, first_output_ms),
+        let (failure, alpha, first_output_ms) = match feedback {
+            AccountAttemptFeedback::Succeeded { first_output_ms } => {
+                (0.0, ACCOUNT_FEEDBACK_EWMA_ALPHA, first_output_ms)
+            }
+            AccountAttemptFeedback::Failed { first_output_ms } => {
+                (1.0, ACCOUNT_FEEDBACK_EWMA_ALPHA, first_output_ms)
+            }
+            AccountAttemptFeedback::CapacityRejected { first_output_ms } => {
+                (1.0, ACCOUNT_CAPACITY_FAILURE_EWMA_ALPHA, first_output_ms)
+            }
         };
         self.failure_rate
             .lock()
             .unwrap_or_else(std::sync::PoisonError::into_inner)
-            .report_at(failure, now);
+            .report_at(failure, alpha, now);
         if let Some(first_output_ms) = first_output_ms.filter(|value| *value > 0) {
             update_feedback_ewma(&self.first_output_ms, first_output_ms as f64);
         }

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

@@ -264,13 +264,17 @@ impl ProviderStreamAccountFeedback {
         {
             return;
         }
-        self.stats.report(
-            &self.provider_kind,
-            &self.account_id,
+        let feedback = if error.kind() == ProviderErrorKind::UpstreamCapacityUnavailable {
+            AccountAttemptFeedback::CapacityRejected {
+                first_output_ms: self.first_output_ms,
+            }
+        } else {
             AccountAttemptFeedback::Failed {
                 first_output_ms: self.first_output_ms,
-            },
-        );
+            }
+        };
+        self.stats
+            .report(&self.provider_kind, &self.account_id, feedback);
         self.reported = true;
     }
 }

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

@@ -537,7 +537,6 @@ impl ProviderCandidate {
 #[derive(Debug, Clone)]
 pub struct RoutingPlan {
     pricing: Arc<crate::metering::PricingOverrides>,
-    disable_fast: bool,
     request_location: Option<crate::account::RequestLocation>,
     config_revision: ConfigRevision,
     account_selection_policy: AccountSelectionPolicy,
@@ -554,8 +553,8 @@ impl RoutingPlan {
     }
 
     #[must_use]
-    pub const fn disable_fast(&self) -> bool {
-        self.disable_fast
+    pub fn disable_fast(&self) -> bool {
+        self.account_scope.disable_fast()
     }
 
     /// 本次请求冻结的全局位置，重试时沿用同一份配置。

backend/crates/gateway-core/src/routing/snapshot.rs

@@ -30,7 +30,6 @@ const MAXIMUM_CATALOG_STABILITY_ATTEMPTS: usize = 4;
 pub struct SnapshotSettingsFacts {
     pricing: Arc<crate::metering::PricingOverrides>,
     request_profiles: BTreeMap<ProviderKind, crate::account::OpaqueProviderData>,
-    disable_fast: bool,
     request_location_enabled: bool,
     request_location: crate::account::RequestLocation,
     max_concurrent_per_account: u32,
@@ -61,12 +60,6 @@ impl SnapshotSettingsFacts {
         self
     }
 
-    #[must_use]
-    pub const fn with_disable_fast(mut self, disable_fast: bool) -> Self {
-        self.disable_fast = disable_fast;
-        self
-    }
-
     #[must_use]
     pub const fn with_responses_max_decompressed_body_bytes(mut self, bytes: u64) -> Self {
         self.responses_max_decompressed_body_bytes = bytes;
@@ -107,7 +100,6 @@ impl SnapshotSettingsFacts {
         min_codex_cli_version: Option<String>,
     ) -> Self {
         Self {
-            disable_fast: false,
             request_profiles: BTreeMap::new(),
             pricing: Arc::default(),
             request_location_enabled: false,
@@ -565,7 +557,6 @@ async fn compile_runtime_snapshot(
     .map(|snapshot| {
         snapshot
             .with_pricing(facts.settings.pricing)
-            .with_disable_fast(facts.settings.disable_fast)
             .with_request_location(request_location)
             .with_responses_max_decompressed_body_bytes(decompressed_body_limit)
             .with_client_queue_policy(client_queue_policy)
@@ -580,7 +571,6 @@ async fn compile_runtime_snapshot(
 #[derive(Debug, Clone)]
 pub struct RuntimeSnapshot {
     pricing: Arc<crate::metering::PricingOverrides>,
-    disable_fast: bool,
     responses_max_decompressed_body_bytes: std::num::NonZeroUsize,
     request_location: Option<crate::account::RequestLocation>,
     revision: ConfigRevision,
@@ -605,12 +595,6 @@ impl RuntimeSnapshot {
         self
     }
 
-    #[must_use]
-    pub const fn with_disable_fast(mut self, disable_fast: bool) -> Self {
-        self.disable_fast = disable_fast;
-        self
-    }
-
     #[must_use]
     pub const fn responses_max_decompressed_body_bytes(&self) -> usize {
         self.responses_max_decompressed_body_bytes.get()
@@ -716,7 +700,6 @@ impl RuntimeSnapshot {
             responses_max_decompressed_body_bytes: std::num::NonZeroUsize::new(64 * 1024 * 1024)
                 .expect("positive default limit"),
             pricing: Arc::default(),
-            disable_fast: false,
             request_location: None,
             revision,
             account_selection_policy,
@@ -1029,7 +1012,6 @@ impl RuntimeSnapshot {
         Ok(RoutingPlan {
             config_revision: self.revision,
             pricing: Arc::clone(&self.pricing),
-            disable_fast: self.disable_fast || account_scope.disable_fast(),
             request_location: self.request_location.clone(),
             account_selection_policy: self.account_selection_policy,
             operation: operation.kind(),
@@ -1073,7 +1055,6 @@ impl RuntimeSnapshot {
         Ok(RoutingPlan {
             config_revision: self.revision,
             pricing: Arc::clone(&self.pricing),
-            disable_fast: self.disable_fast || account_scope.disable_fast(),
             request_location: self.request_location.clone(),
             account_selection_policy: self.account_selection_policy,
             operation: operation.kind(),

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

@@ -34,6 +34,58 @@ fn report_failure(
     );
 }
 
+#[test]
+fn capacity_rejections_should_raise_failure_rate_faster_than_regular_failures() {
+    let (feedback, provider, account) = feedback_subject();
+    let observed_at = Instant::now();
+    let mut failure_rates = Vec::new();
+    for _ in 0..2 {
+        feedback.report_at(
+            &provider,
+            &account,
+            AccountAttemptFeedback::CapacityRejected {
+                first_output_ms: None,
+            },
+            observed_at,
+        );
+        failure_rates.push(
+            feedback
+                .scheduling_signals_at(&provider, &account, observed_at)
+                .0,
+        );
+    }
+
+    assert_eq!(failure_rates, [Some(4_000), Some(6_400)]);
+}
+
+#[test]
+fn capacity_failure_rate_should_keep_time_decay_and_success_recovery() {
+    let (feedback, provider, account) = feedback_subject();
+    let observed_at = Instant::now();
+    feedback.report_at(
+        &provider,
+        &account,
+        AccountAttemptFeedback::CapacityRejected {
+            first_output_ms: Some(100),
+        },
+        observed_at,
+    );
+    let recovered_at = observed_at + FAILURE_RATE_HALF_LIFE;
+    feedback.report_at(
+        &provider,
+        &account,
+        AccountAttemptFeedback::Succeeded {
+            first_output_ms: Some(200),
+        },
+        recovered_at,
+    );
+
+    assert_eq!(
+        feedback.scheduling_signals_at(&provider, &account, recovered_at),
+        (Some(1_600), Some(120)),
+    );
+}
+
 #[test]
 fn account_failure_rate_should_halve_after_one_half_life() {
     let (feedback, provider, account) = feedback_subject();

backend/crates/gateway-core/tests/engine/coordinator.rs

@@ -677,7 +677,6 @@ fn plan_with_profiles(
         Vec::new(),
     )
     .expect("snapshot")
-    .with_disable_fast(disable_fast)
     .with_request_location(Some(request_location))
     .with_account_directory(Arc::clone(&directory));
     snapshot
@@ -686,6 +685,7 @@ fn plan_with_profiles(
             operation,
             Arc::new(
                 FrozenAccountScope::new(directory, ClientRoutingScope::all_accounts())
+                    .with_disable_fast(disable_fast)
                     .with_request_profiles(profiles),
             ),
             &RoutingContext {
@@ -4570,7 +4570,7 @@ fn deadline_before_first_event_records_no_provider_circuit_failure() {
 }
 
 #[test]
-fn global_location_and_fast_policy_reach_every_account_retry() {
+fn global_location_and_group_fast_policy_reach_every_account_retry() {
     let operation = generate_operation();
     let location = gateway_core::account::RequestLocation {
         timezone: "Asia/Tokyo".parse().unwrap(),

backend/crates/gateway-core/tests/engine/execution.rs

@@ -178,7 +178,7 @@ fn reused_client_uses_updated_limits_for_each_execution() {
         requests_per_minute: 1,
     };
     for (revision, limits) in [(2, limited), (3, RateLimits::unlimited())] {
-        snapshots.publish(start_snapshot_with_policy(revision, true, limits));
+        snapshots.publish(start_snapshot_with_policy(revision, true, limits, false));
         let mut next = request(&service, ClientTransport::WebSocket);
         next.client = client.clone();
         let started = block_on(service.start(next)).expect("new execution");
@@ -212,6 +212,7 @@ fn reused_client_cannot_start_after_key_disable_or_snapshot_suspension() {
                 2,
                 false,
                 RateLimits::unlimited(),
+                false,
             ));
         }
         let result = block_on(service.start(next));
@@ -1893,10 +1894,15 @@ fn client_snapshot() -> RuntimeSnapshot {
 }
 
 fn start_snapshot() -> RuntimeSnapshot {
-    start_snapshot_with_policy(1, true, RateLimits::unlimited())
+    start_snapshot_with_policy(1, true, RateLimits::unlimited(), false)
 }
 
-fn start_snapshot_with_policy(revision: u64, enabled: bool, limits: RateLimits) -> RuntimeSnapshot {
+fn start_snapshot_with_policy(
+    revision: u64,
+    enabled: bool,
+    limits: RateLimits,
+    disable_fast: bool,
+) -> RuntimeSnapshot {
     let provider = ProviderKind::new("openai").expect("provider kind");
     let capabilities =
         ModelCapabilities::new(BTreeSet::from([OperationKind::Generate]), Some(16_000));
@@ -1916,7 +1922,12 @@ fn start_snapshot_with_policy(revision: u64, enabled: bool, limits: RateLimits)
         vec![ClientPolicy::new(
             ClientApiKeyId::new("key_start_test").expect("client API key ID"),
             PlaintextClientApiKey::new("sk_start_test").expect("plaintext client API key"),
-            account_scope(&provider, "acct_start"),
+            Arc::new(
+                account_scope(&provider, "acct_start")
+                    .as_ref()
+                    .clone()
+                    .with_disable_fast(disable_fast),
+            ),
             enabled,
             limits,
         )],
@@ -2013,6 +2024,7 @@ fn account_wait_inherits_the_budget_spent_during_client_admission() {
                 max_concurrency: 1,
                 requests_per_minute: 0,
             },
+            false,
         )
         .with_client_queue_policy(gateway_core::concurrency::ConcurrencyQueuePolicy {
             max_waiting: 1,
@@ -2123,6 +2135,7 @@ fn queue_service(
             max_concurrency,
             requests_per_minute: 0,
         },
+        false,
     )
     .with_client_queue_policy(gateway_core::concurrency::ConcurrencyQueuePolicy {
         max_waiting,
@@ -2249,7 +2262,7 @@ fn cancelling_pending_admission_cleans_up_a_slot_acquired_before_the_reply() {
 }
 
 #[test]
-fn reused_websocket_client_gets_fast_policy_from_each_new_request_snapshot() {
+fn reused_websocket_client_gets_group_fast_policy_from_each_new_request_snapshot() {
     let snapshots = RuntimeSnapshotHandle::new(start_snapshot());
     let admissions = Arc::new(Admissions::default());
     let provider = Arc::new(ChargedProvider::default());
@@ -2264,10 +2277,12 @@ fn reused_websocket_client_gets_fast_policy_from_each_new_request_snapshot() {
     );
     let client = service.authenticate("sk_start_test").unwrap();
     for (revision, disable_fast) in [(2, true), (3, false)] {
-        snapshots.publish(
-            start_snapshot_with_policy(revision, true, RateLimits::unlimited())
-                .with_disable_fast(disable_fast),
-        );
+        snapshots.publish(start_snapshot_with_policy(
+            revision,
+            true,
+            RateLimits::unlimited(),
+            disable_fast,
+        ));
         let mut next = request(&service, ClientTransport::WebSocket);
         next.client = client.clone();
         let mut started = block_on(service.start(next)).unwrap();

backend/crates/gateway-core/tests/routing/snapshot.rs

@@ -494,10 +494,10 @@ fn decompression_setting_should_validate_and_remain_frozen_across_publication()
 }
 
 #[test]
-fn disable_fast_uses_bound_groups_and_global_policy_without_changing_account_scope() {
+fn disable_fast_uses_only_bound_groups_without_changing_account_scope() {
     use gateway_core::account::ProviderAccountId;
     use gateway_core::routing::AccountGroupId;
-    for global in [false, true] {
+    for disable_fast in [false, true] {
         for group_enabled in [false, true] {
             for bound in [false, true] {
                 let group_id = AccountGroupId::new("grp_00000000000000000000000000000001").unwrap();
@@ -507,8 +507,7 @@ fn disable_fast_uses_bound_groups_and_global_policy_without_changing_account_sco
                 let facts = SnapshotFacts::new(
                     revision(1),
                     revision(1),
-                    SnapshotSettingsFacts::new(3, 0, "smart", BTreeMap::new(), None, None)
-                        .with_disable_fast(global),
+                    SnapshotSettingsFacts::new(3, 0, "smart", BTreeMap::new(), None, None),
                     vec![SnapshotClientPolicyFacts::new(
                         ClientApiKeyId::new("key_fast_policy").unwrap(),
                         PlaintextClientApiKey::new("sk_fast_policy").unwrap(),
@@ -525,7 +524,7 @@ fn disable_fast_uses_bound_groups_and_global_policy_without_changing_account_sco
                             "Restricted".to_owned(),
                             group_enabled,
                         )
-                        .with_disable_fast(true),
+                        .with_disable_fast(disable_fast),
                         SnapshotAccountGroupFacts::new(
                             open_group_id.clone(),
                             "Open".to_owned(),
@@ -566,8 +565,8 @@ fn disable_fast_uses_bound_groups_and_global_policy_without_changing_account_sco
                     .unwrap();
                 assert_eq!(
                     plan.disable_fast(),
-                    global || bound,
-                    "global={global}, enabled={group_enabled}, bound={bound}"
+                    disable_fast && bound,
+                    "disable_fast={disable_fast}, enabled={group_enabled}, bound={bound}"
                 );
             }
         }

backend/crates/gateway-store/src/admin_adapter.rs

@@ -24,7 +24,7 @@ impl SettingsStore for AdminSettingsStoreAdapter {
 
     async fn sync_pricing(
         &self,
-        prices: gateway_core::metering::PricingOverrides,
+        changes: gateway_admin::model::pricing::PricingSyncChanges,
         context: &MutationContext,
     ) -> AdminStoreResult<gateway_admin::model::Revision> {
         let audit = mutation_audit(
@@ -36,7 +36,7 @@ impl SettingsStore for AdminSettingsStoreAdapter {
         );
         let revision = self
             .control_plane
-            .sync_pricing(prices, audit)
+            .sync_pricing(changes, audit)
             .await
             .map_err(|error| admin_store_error("model pricing sync", error))?;
         admin_revision(revision)
@@ -91,7 +91,6 @@ impl SettingsStore for AdminSettingsStoreAdapter {
                 refresh_margin_seconds: command.refresh_margin_seconds,
                 refresh_concurrency: command.refresh_concurrency,
                 max_concurrent_per_account: command.max_concurrent_per_account,
-                disable_fast: command.disable_fast,
                 request_location_enabled: command.request_location_enabled,
                 request_location: command.request_location,
                 request_interval_ms: command.request_interval_ms,
@@ -123,7 +122,6 @@ impl SettingsStore for AdminSettingsStoreAdapter {
                 "1",
                 vec![
                     "provider_request_profiles_json".to_owned(),
-                    "disable_fast".to_owned(),
                     "request_location_enabled".to_owned(),
                     "request_location_json".to_owned(),
                     "model_mappings_json".to_owned(),
@@ -235,7 +233,6 @@ pub(crate) fn admin_runtime_settings(
     Ok(AdminRuntimeSettings {
         openai_client_profile: settings.openai_client_profile,
         config_revision: admin_revision(settings.config_revision)?,
-        disable_fast: settings.disable_fast,
         request_location_enabled: settings.request_location_enabled,
         request_location: settings.request_location,
         model_mappings,

backend/crates/gateway-store/src/postgres/pricing.rs

@@ -1,6 +1,6 @@
 //! 价格覆盖与请求费用明细的持久化。
 
-use gateway_admin::model::pricing::{PricingChange, UpdatePricing};
+use gateway_admin::model::pricing::{PricingChange, PricingSyncChanges, UpdatePricing};
 use gateway_core::metering::{ModelPriceOverride, PricingOverrides};
 use sqlx::types::Json;
 
@@ -146,16 +146,32 @@ impl PgControlPlaneRepository {
 
     pub(crate) async fn sync_pricing(
         &self,
-        prices: PricingOverrides,
+        changes: PricingSyncChanges,
         audit: AdminAuditEvent,
     ) -> StoreResult<Revision> {
-        validate_pricing(&prices)?;
         let mut transaction = self
             .pool
             .begin()
             .await
             .map_err(|_| postgres_unavailable("begin pricing sync"))?;
-        // 同步只替换来源层；不写人工覆盖。运行设置行锁统一串行化配置修订。
+        let Json(mut prices) = sqlx::query_scalar::<_, Json<PricingOverrides>>(
+            "select pricing_synced_json from runtime_settings where id = 1 for update",
+        )
+        .fetch_one(&mut *transaction)
+        .await
+        .map_err(|_| postgres_unavailable("lock synced model pricing"))?;
+        for (provider, models) in changes {
+            let stored = prices.entry(provider).or_default();
+            for (model, price) in models {
+                if let Some(price) = price {
+                    stored.insert(model, price);
+                } else {
+                    stored.remove(&model);
+                }
+            }
+        }
+        prices.retain(|_, models| !models.is_empty());
+        validate_pricing(&prices)?;
         sqlx::query("update runtime_settings set pricing_synced_json = $1, pricing_synced_at = now() where id = 1")
             .bind(Json(prices)).execute(&mut *transaction).await.map_err(|_| postgres_unavailable("sync model pricing"))?;
         let revision = bump_config_revision_in_transaction(&mut transaction).await?;
@@ -178,18 +194,23 @@ impl PgControlPlaneRepository {
             .await
             .map_err(|_| postgres_unavailable("begin model pricing update"))?;
         // 锁定当前配置再更新选中项，避免批量操作覆盖其他管理员已提交的模型。
-        let Json(mut pricing) = sqlx::query_scalar::<_, Json<PricingOverrides>>(
-            "select pricing_overrides_json from runtime_settings where id = 1 for update",
+        let (Json(mut pricing), Json(mut synced)) = sqlx::query_as::<_, (Json<PricingOverrides>, Json<PricingOverrides>)>(
+            "select pricing_overrides_json, pricing_synced_json from runtime_settings where id = 1 for update",
         )
         .fetch_one(&mut *transaction)
         .await
         .map_err(|_| postgres_unavailable("lock model pricing"))?;
-        let models = pricing.entry(command.provider).or_default();
+        let models = pricing.entry(command.provider.clone()).or_default();
+        let source = synced.entry(command.provider).or_default();
         for model in command.models {
             match &command.change {
                 PricingChange::Reset => {
                     models.remove(&model);
                 }
+                PricingChange::Delete => {
+                    models.remove(&model);
+                    source.remove(&model);
+                }
                 PricingChange::Replace(value) => {
                     models.insert(model, value.clone());
                 }
@@ -205,9 +226,11 @@ impl PgControlPlaneRepository {
             }
         }
         pricing.retain(|_, models| !models.is_empty());
+        synced.retain(|_, models| !models.is_empty());
         validate_pricing(&pricing)?;
-        sqlx::query("update runtime_settings set pricing_overrides_json = $1 where id = 1")
+        sqlx::query("update runtime_settings set pricing_overrides_json = $1, pricing_synced_json = $2 where id = 1")
             .bind(Json(pricing))
+            .bind(Json(synced))
             .execute(&mut *transaction)
             .await
             .map_err(|_| postgres_unavailable("update model pricing"))?;

backend/crates/gateway-store/src/postgres/runtime_settings.rs

@@ -20,7 +20,6 @@ use crate::{Revision, StoreError, StoreResult, postgres_unavailable};
 #[derive(Clone, PartialEq, Eq)]
 pub struct RuntimeSettings {
     pub openai_client_profile: Option<gateway_core::account::OpaqueProviderData>,
-    pub disable_fast: bool,
     pub config_revision: Revision,
     pub admin_api_key: Option<String>,
     pub refresh_margin_seconds: u64,
@@ -66,7 +65,6 @@ impl fmt::Debug for RuntimeSettings {
                 &self.max_concurrent_per_account,
             )
             .field("request_interval_ms", &self.request_interval_ms)
-            .field("disable_fast", &self.disable_fast)
             .field("rotation_strategy", &self.rotation_strategy)
             .field("request_location_enabled", &self.request_location_enabled)
             .field("request_location", &self.request_location)
@@ -112,7 +110,6 @@ impl fmt::Debug for RuntimeSettings {
 #[derive(Clone)]
 pub struct RuntimeSettingsUpdate {
     pub openai_client_profile: Option<gateway_core::account::OpaqueProviderData>,
-    pub disable_fast: Option<bool>,
     pub admin_api_key: Option<String>,
     pub refresh_margin_seconds: u64,
     pub refresh_concurrency: u32,
@@ -148,7 +145,6 @@ impl fmt::Debug for RuntimeSettingsUpdate {
                 "admin_api_key",
                 &self.admin_api_key.as_ref().map(|_| "[REDACTED]"),
             )
-            .field("disable_fast", &self.disable_fast)
             .field("rotation_strategy", &self.rotation_strategy)
             .field("request_location_enabled", &self.request_location_enabled)
             .field("request_location", &self.request_location)
@@ -235,7 +231,7 @@ impl RuntimeSettingsRepository for PgRuntimeSettingsRepository {
 
 pub(crate) async fn load_runtime_settings_from_pool(pool: &PgPool) -> StoreResult<RuntimeSettings> {
     let row = sqlx::query_as::<_, RuntimeSettingsRow>(
-            "select provider_request_profiles_json, config_revision, admin_api_key, refresh_margin_seconds, request_location_json, request_location_enabled, disable_fast,
+            "select provider_request_profiles_json, config_revision, admin_api_key, refresh_margin_seconds, request_location_json, request_location_enabled,
                     refresh_concurrency, max_concurrent_per_account, request_interval_ms,
                     rotation_strategy, model_mappings_json, usage_retention_days, ops_event_retention_days,
                     audit_retention_days, min_codex_desktop_version,
@@ -321,7 +317,7 @@ pub(crate) async fn load_runtime_settings_in_transaction(
     transaction: &mut Transaction<'_, Postgres>,
 ) -> StoreResult<RuntimeSettings> {
     let row = sqlx::query_as::<_, RuntimeSettingsRow>(
-        "select provider_request_profiles_json, config_revision, admin_api_key, refresh_margin_seconds, request_location_json, request_location_enabled, disable_fast,
+        "select provider_request_profiles_json, config_revision, admin_api_key, refresh_margin_seconds, request_location_json, request_location_enabled,
                 refresh_concurrency, max_concurrent_per_account, request_interval_ms,
                 rotation_strategy, model_mappings_json, usage_retention_days, ops_event_retention_days,
                 audit_retention_days, min_codex_desktop_version,
@@ -377,8 +373,7 @@ pub(crate) async fn update_runtime_settings_in_transaction(
                      request_location_json = $23,
                      request_location_enabled = $24,
                      responses_max_decompressed_body_bytes = $25,
-                     disable_fast = coalesce($26, disable_fast),
-                     provider_request_profiles_json = case when $27::jsonb is null then provider_request_profiles_json else jsonb_set(provider_request_profiles_json, '{openai}', $27) end,
+                     provider_request_profiles_json = case when $26::jsonb is null then provider_request_profiles_json else jsonb_set(provider_request_profiles_json, '{openai}', $26) end,
 	                 updated_at = now()
 	             where id = 1
 	             returning config_revision",
@@ -420,7 +415,6 @@ pub(crate) async fn update_runtime_settings_in_transaction(
         i64::try_from(update.responses_max_decompressed_body_bytes)
             .map_err(|_| invalid_numeric())?,
     )
-    .bind(update.disable_fast)
     .bind(update.openai_client_profile.as_ref().map(|profile| sqlx::types::Json(profile.expose_to_provider())))
     .fetch_optional(&mut **transaction)
     .await
@@ -474,7 +468,6 @@ struct RuntimeSettingsRow {
     provider_request_profiles_json: sqlx::types::Json<
         std::collections::BTreeMap<String, serde_json::Map<String, serde_json::Value>>,
     >,
-    disable_fast: bool,
     config_revision: i64,
     admin_api_key: Option<String>,
     refresh_margin_seconds: i64,
@@ -518,7 +511,6 @@ fn runtime_settings_from_row(mut row: RuntimeSettingsRow) -> StoreResult<Runtime
         max_concurrent_per_account: to_u32(row.max_concurrent_per_account)?,
         request_interval_ms: to_u64(row.request_interval_ms)?,
         rotation_strategy: row.rotation_strategy,
-        disable_fast: row.disable_fast,
         request_location_enabled: row.request_location_enabled,
         request_location: row
             .request_location_json

backend/crates/gateway-store/src/postgres/snapshot.rs

@@ -23,7 +23,6 @@ pub struct SnapshotRuntimeSettings {
     pub pricing: gateway_core::metering::PricingOverrides,
     pub request_profiles:
         BTreeMap<gateway_core::routing::ProviderKind, gateway_core::account::OpaqueProviderData>,
-    pub disable_fast: bool,
     pub request_location_enabled: bool,
     pub request_location: gateway_core::account::RequestLocation,
     pub refresh_margin_seconds: u64,
@@ -163,7 +162,6 @@ impl SnapshotStorePort for PgRuntimeSnapshotRepository {
             .with_responses_max_decompressed_body_bytes(
                 data.settings.responses_max_decompressed_body_bytes,
             )
-            .with_disable_fast(data.settings.disable_fast)
             .with_request_profiles(data.settings.request_profiles)
             .with_pricing(data.settings.pricing)
             .with_request_location(
@@ -266,7 +264,6 @@ struct SnapshotSettingsRow {
     request_location_json: sqlx::types::Json<gateway_core::account::RequestLocation>,
     request_location_enabled: bool,
     responses_max_decompressed_body_bytes: i64,
-    disable_fast: bool,
     provider_request_profiles_json:
         sqlx::types::Json<BTreeMap<String, serde_json::Map<String, serde_json::Value>>>,
 }
@@ -275,7 +272,7 @@ async fn load_settings(
     transaction: &mut Transaction<'_, Postgres>,
 ) -> StoreResult<(Revision, SnapshotRuntimeSettings)> {
     let row = sqlx::query_as::<_, SnapshotSettingsRow>(
-        "select config_revision, refresh_margin_seconds, refresh_concurrency, max_concurrent_per_account, request_interval_ms, rotation_strategy, model_mappings_json, min_codex_desktop_version, min_codex_cli_version, max_waiting_per_key, max_waiting_per_account, concurrency_wait_timeout_seconds, request_location_json, request_location_enabled, responses_max_decompressed_body_bytes, disable_fast, provider_request_profiles_json, pricing_overrides_json, pricing_synced_json from runtime_settings where id = 1",
+        "select config_revision, refresh_margin_seconds, refresh_concurrency, max_concurrent_per_account, request_interval_ms, rotation_strategy, model_mappings_json, min_codex_desktop_version, min_codex_cli_version, max_waiting_per_key, max_waiting_per_account, concurrency_wait_timeout_seconds, request_location_json, request_location_enabled, responses_max_decompressed_body_bytes, provider_request_profiles_json, pricing_overrides_json, pricing_synced_json from runtime_settings where id = 1",
     )
     .fetch_optional(&mut **transaction)
     .await
@@ -296,7 +293,6 @@ async fn load_settings(
                 )
             },
             request_profiles: decode_request_profiles(row.provider_request_profiles_json.0)?,
-            disable_fast: row.disable_fast,
             responses_max_decompressed_body_bytes: to_u64(
                 row.responses_max_decompressed_body_bytes,
             )?,

backend/crates/gateway-store/tests/postgres/pricing.rs

@@ -8,14 +8,9 @@ use serde_json::json;
 
 use super::TestDatabase;
 
-#[tokio::test]
-async fn concurrent_price_edits_preserve_other_models_and_sync_preserves_manual_prices() {
-    let Some(redis_url) = crate::support::test_env("CPR_TEST_REDIS_URL") else {
-        return;
-    };
-    let Some(database) = TestDatabase::create("pricing").await else {
-        return;
-    };
+async fn pricing_store(name: &str) -> Option<(TestDatabase, gateway_store::StoreBundle)> {
+    let redis_url = crate::support::test_env("CPR_TEST_REDIS_URL")?;
+    let database = TestDatabase::create(name).await?;
     let mut database_url = url::Url::parse(
         &crate::support::test_env("CPR_TEST_DATABASE_URL").expect("test database URL"),
     )
@@ -36,6 +31,14 @@ async fn concurrent_price_edits_preserve_other_models_and_sync_preserves_manual_
     let runtime = tempfile::tempdir().unwrap();
     config.resolve_and_validate(runtime.path()).unwrap();
     let bundle = gateway_store::initialize(config).await.unwrap();
+    Some((database, bundle))
+}
+
+#[tokio::test]
+async fn concurrent_price_edits_preserve_other_models_and_sync_preserves_manual_prices() {
+    let Some((database, bundle)) = pricing_store("pricing").await else {
+        return;
+    };
     let settings = bundle.admin_ports().settings();
     let context = MutationContext {
         actor: MutationActor::System,
@@ -76,7 +79,21 @@ async fn concurrent_price_edits_preserve_other_models_and_sync_preserves_manual_
         .unwrap()
         .input = "9".to_owned().try_into().unwrap();
     settings
-        .sync_pricing(source.clone(), &context)
+        .sync_pricing(
+            source
+                .iter()
+                .map(|(provider, models)| {
+                    (
+                        provider.clone(),
+                        models
+                            .iter()
+                            .map(|(model, price)| (model.clone(), Some(price.clone())))
+                            .collect(),
+                    )
+                })
+                .collect(),
+            &context,
+        )
         .await
         .unwrap();
     let stored = settings.load_pricing().await.unwrap();
@@ -123,3 +140,154 @@ async fn concurrent_price_edits_preserve_other_models_and_sync_preserves_manual_
     drop(bundle);
     database.close().await;
 }
+
+#[tokio::test]
+async fn deleting_pricing_removes_both_layers_and_preserves_other_models_and_frozen_snapshots() {
+    let Some((database, bundle)) = pricing_store("pricing_delete").await else {
+        return;
+    };
+    let settings = bundle.admin_ports().settings();
+    let context = MutationContext {
+        actor: MutationActor::System,
+        request_id: "pricing-delete-test".to_owned(),
+    };
+    let value = json!({
+        "multiplierBps":10000,
+        "bands":{"standard":{"input":"2","output":"10","cacheRead":"0","cacheWrite":"1"}}
+    });
+    settings
+        .sync_pricing(
+            serde_json::from_value(json!({
+                "openai":{"shared":value,"synced-only":value,"untouched":value},
+                "xai":{"shared":value}
+            }))
+            .unwrap(),
+            &context,
+        )
+        .await
+        .unwrap();
+    settings
+        .update_pricing(
+            UpdatePricing {
+                provider: "openai".to_owned(),
+                models: vec![
+                    "shared".to_owned(),
+                    "custom-only".to_owned(),
+                    "untouched".to_owned(),
+                ],
+                change: PricingChange::Replace(serde_json::from_value(value.clone()).unwrap()),
+            },
+            &context,
+        )
+        .await
+        .unwrap();
+    let repository = PgRuntimeSnapshotRepository::new(database.pool.clone());
+    let frozen = repository.load_runtime_snapshot().await.unwrap();
+    let synced_at = settings.load_pricing().await.unwrap().synced_at;
+    settings
+        .update_pricing(
+            UpdatePricing {
+                provider: "openai".to_owned(),
+                models: vec![
+                    "shared".to_owned(),
+                    "synced-only".to_owned(),
+                    "custom-only".to_owned(),
+                ],
+                change: PricingChange::Delete,
+            },
+            &context,
+        )
+        .await
+        .unwrap();
+    let stored = settings.load_pricing().await.unwrap();
+    assert_eq!(
+        serde_json::to_value(&stored.overrides).unwrap(),
+        json!({"openai":{"untouched":value}})
+    );
+    assert_eq!(
+        serde_json::to_value(&stored.synced).unwrap(),
+        json!({"openai":{"untouched":value},"xai":{"shared":value}})
+    );
+    assert_eq!(stored.synced_at, synced_at);
+    let current = repository.load_runtime_snapshot().await.unwrap();
+    assert_eq!(current.settings.pricing["openai"].len(), 1);
+    assert!(current.settings.pricing["xai"].contains_key("shared"));
+    assert_eq!(frozen.settings.pricing["openai"].len(), 4);
+    let audits: i64 = sqlx::query_scalar(
+        "select count(*) from admin_audit_events where action = 'pricing.update'",
+    )
+    .fetch_one(&database.pool)
+    .await
+    .unwrap();
+    assert_eq!(audits, 2);
+    drop(settings);
+    drop(bundle);
+    database.close().await;
+}
+
+#[tokio::test]
+async fn concurrent_pricing_syncs_merge_selected_changes_and_remove_only_selected_models() {
+    let Some((database, bundle)) = pricing_store("pricing_selection").await else {
+        return;
+    };
+    let settings = bundle.admin_ports().settings();
+    let context = MutationContext {
+        actor: MutationActor::System,
+        request_id: "pricing-selection-test".to_owned(),
+    };
+    let price = json!({
+        "multiplierBps":10000,
+        "bands":{"standard":{"input":"2","output":"10","cacheRead":"0","cacheWrite":"1"}}
+    });
+    let patch = |value| {
+        serde_json::from_value::<gateway_admin::model::pricing::PricingSyncChanges>(value).unwrap()
+    };
+    let (first, second) = tokio::join!(
+        settings.sync_pricing(patch(json!({"openai": {"model-a": price}})), &context),
+        settings.sync_pricing(
+            patch(json!({"openai": {"model-b": price}, "xai": {"model-a": price}})),
+            &context
+        ),
+    );
+    assert_ne!(first.unwrap(), second.unwrap());
+    let stored = settings.load_pricing().await.unwrap();
+    assert_eq!(
+        serde_json::to_value(&stored.synced).unwrap(),
+        json!({
+            "openai": {"model-a": price, "model-b": price}, "xai": {"model-a": price}
+        })
+    );
+    settings
+        .sync_pricing(patch(json!({"openai": {"model-a": null}})), &context)
+        .await
+        .unwrap();
+    let stored = settings.load_pricing().await.unwrap();
+    assert_eq!(
+        serde_json::to_value(&stored.synced).unwrap(),
+        json!({
+            "openai": {"model-b": price}, "xai": {"model-a": price}
+        })
+    );
+    let before = stored;
+    assert!(
+        settings
+            .sync_pricing(
+                patch(json!({
+                    "openai": {"model-b": null}, "invalid": {"model-c": price}
+                })),
+                &context
+            )
+            .await
+            .is_err()
+    );
+    assert_eq!(settings.load_pricing().await.unwrap(), before);
+    let audits: i64 =
+        sqlx::query_scalar("select count(*) from admin_audit_events where action = 'pricing.sync'")
+            .fetch_one(&database.pool)
+            .await
+            .unwrap();
+    assert_eq!(audits, 3);
+    drop(settings);
+    drop(bundle);
+    database.close().await;
+}

backend/crates/gateway-store/tests/postgres/runtime_settings.rs

@@ -10,7 +10,6 @@ use super::TestDatabase;
 fn settings_with_margin(refresh_margin_seconds: u64) -> RuntimeSettingsUpdate {
     RuntimeSettingsUpdate {
         openai_client_profile: None,
-        disable_fast: None,
         request_location_enabled: false,
         request_location: Default::default(),
         admin_api_key: None,
@@ -432,32 +431,6 @@ async fn decompression_setting_should_persist_and_reach_snapshot_facts() {
     database.close().await;
 }
 
-#[tokio::test]
-async fn disable_fast_persists_and_omitted_updates_preserve_the_restriction() {
-    let Some(database) = TestDatabase::create("disable_fast_settings").await else {
-        return;
-    };
-    let repository = PgRuntimeSettingsRepository::new(database.pool.clone());
-    let initial = repository.load_runtime_settings().await.unwrap();
-    assert!(!initial.disable_fast);
-    for (value, expected) in [(Some(true), true), (None, true), (Some(false), false)] {
-        let mut update = settings_with_margin(1_800);
-        update.disable_fast = value;
-        let revision = repository.update_runtime_settings(update).await.unwrap();
-        let reloaded = repository.load_runtime_settings().await.unwrap();
-        assert_eq!(reloaded.disable_fast, expected);
-        assert_eq!(reloaded.config_revision, revision);
-        use gateway_store::postgres::{PgRuntimeSnapshotRepository, RuntimeSnapshotRepository};
-        let snapshot = PgRuntimeSnapshotRepository::new(database.pool.clone())
-            .load_runtime_snapshot()
-            .await
-            .unwrap();
-        assert_eq!(snapshot.settings.disable_fast, expected);
-        assert_eq!(snapshot.config_revision, revision);
-    }
-    database.close().await;
-}
-
 #[tokio::test]
 async fn request_profile_initialization_is_idempotent_and_old_updates_preserve_it() {
     use gateway_core::{

backend/crates/providers/openai/tests/provider/failure.rs

@@ -210,12 +210,47 @@ fn openai_feedback_should_penalize_capacity_rejections_even_with_transient_retry
 
         deliver_error_with_openai_feedback(&feedback, &provider, &account, error);
 
-        assert!(
-            feedback
-                .scheduling_signals(&provider, &account)
-                .0
-                .is_some_and(|failure_rate| failure_rate > 0),
-            "capacity rejection did not affect the account score: {code}"
+        assert_eq!(
+            feedback.scheduling_signals(&provider, &account).0,
+            Some(4_000),
+            "capacity rejection should receive the stronger penalty: {code}"
+        );
+    }
+}
+
+#[test]
+fn openai_feedback_should_keep_regular_server_error_penalty() {
+    let feedback = Arc::new(AccountFeedbackStats::default());
+    let provider = ProviderKind::new("openai").expect("provider");
+    let account = ProviderAccountId::new("acct_server_error").expect("account");
+
+    deliver_error_with_openai_feedback(
+        &feedback,
+        &provider,
+        &account,
+        sent_error(ProviderErrorKind::Unavailable, Some("server_error")),
+    );
+
+    assert_eq!(
+        feedback.scheduling_signals(&provider, &account).0,
+        Some(2_000)
+    );
+}
+
+#[test]
+fn openai_feedback_should_ignore_unconfirmed_capacity_rejections() {
+    for send_state in [UpstreamSendState::NotSent, UpstreamSendState::Ambiguous] {
+        let feedback = Arc::new(AccountFeedbackStats::default());
+        let provider = ProviderKind::new("openai").expect("provider");
+        let account = ProviderAccountId::new("acct_unconfirmed").expect("account");
+        let error = ProviderError::new(ProviderErrorKind::UpstreamCapacityUnavailable, send_state)
+            .with_upstream_code(OpaqueUpstreamValue::new("server_is_overloaded"));
+
+        deliver_error_with_openai_feedback(&feedback, &provider, &account, error);
+
+        assert_eq!(
+            feedback.scheduling_signals(&provider, &account),
+            (None, None)
         );
     }
 }

backend/migrations/.frozen-sha256

@@ -13,5 +13,4 @@ b20e425c0b062b72d35fdebe294e023eb17ceb44ac93083ff3d48400715bf430  0012_upstream_
 6e23fceef1096b50c1af3584cb1687eded2ad37e31e65391397ec984bf78c7e3  0013_responses_decompression_limit.sql
 12a540be31d65c4925cbe63dab42ea2d4ec2864fecf71a6dd4ee942e71726857  0014_disable_fast.sql
 9bd37bd314b6864816c43cb4fd1869fb415920d4f10f766976e344334a1f60fc  0015_proxy_dual_stack_exit.sql
-3eb917aab407a497355b21db0faf7974aac757801473b0a13b47b136598b371f  0016_client_request_profiles.sql
-3569e370689caac3f03e76539bc7df4b22304988f428cb7f8911aba170c59167  0017_model_pricing.sql
+a1e23b8535ed41de27861833cd2d5ee7498a9bb0e523156ba3f011f9c6b3966b  0016_client_profiles_and_model_pricing.sql

backend/migrations/0016_client_profiles_and_model_pricing.sql

@@ -9,3 +9,14 @@ alter table client_api_keys
     add column provider_request_profiles_json jsonb not null default '{}'::jsonb,
     add constraint client_api_keys_request_profiles_object
         check (jsonb_typeof(provider_request_profiles_json) = 'object');
+
+-- 覆盖配置与请求历史明细分离，改价不重算已发生的费用。
+alter table runtime_settings add column pricing_overrides_json jsonb not null default '{}'
+    check (jsonb_typeof(pricing_overrides_json) = 'object');
+alter table runtime_settings add column pricing_synced_json jsonb not null default '{}'
+    check (jsonb_typeof(pricing_synced_json) = 'object');
+alter table runtime_settings add column pricing_synced_at timestamptz;
+alter table model_requests add column billing_snapshot_json jsonb
+    check (billing_snapshot_json is null or jsonb_typeof(billing_snapshot_json) = 'object');
+
+alter table runtime_settings drop column disable_fast;

backend/migrations/0017_model_pricing.sql

@@ -1,8 +0,0 @@
--- 覆盖配置与请求历史明细分离，改价不重算已发生的费用。
-alter table runtime_settings add column pricing_overrides_json jsonb not null default '{}'
-    check (jsonb_typeof(pricing_overrides_json) = 'object');
-alter table runtime_settings add column pricing_synced_json jsonb not null default '{}'
-    check (jsonb_typeof(pricing_synced_json) = 'object');
-alter table runtime_settings add column pricing_synced_at timestamptz;
-alter table model_requests add column billing_snapshot_json jsonb
-    check (billing_snapshot_json is null or jsonb_typeof(billing_snapshot_json) = 'object');

deploy/README.md

@@ -64,7 +64,7 @@ Linux 上应用容器以 `10001:10001` 运行。上述命令将应用数据和
 配置设为 `0640`，均由当前用户持有、容器组 `10001` 访问。
 `config.yaml` 通过 Compose `configs` 只读挂载，普通 Compose 保留宿主机文件的 UID/GID 和 mode。
 
-OpenAI 上游身份在管理端「系统设置 → 上游身份」配置，Key 可选择独立身份覆盖通用值。
+OpenAI 上游身份在管理端「系统设置 → 上游配置 → 客户端身份」配置，Key 可选择独立身份覆盖通用值。
 首次默认 macOS Desktop 自动最新，支持范围与字段见 [客户端身份合同](../docs/api.md#openai-上游客户端身份)。
 旧 `openai.wire_profile` 在数据库尚无选择时一次性导入：默认版本采用自动更新；版本元组、平台或架构
 与默认值不同则保留为固定版本。导入后从管理端修改，重启不会用 YAML 覆盖数据库。
@@ -72,7 +72,7 @@ OpenAI 上游身份在管理端「系统设置 → 上游身份」配置，Key
 OpenAI 上游地址、WebSocket 池、额度刷新和 OAuth 使用 Provider 默认值；xAI 仍在 YAML 配置启动画像。
 后台发布检查只更新可重建的版本资料，不回写用户选择或 `config.yaml`；失败继续使用同组合的有效版本。
 
-全局请求位置在管理端「系统设置 → 运行设置 → 请求时区与位置」配置，默认关闭，保留客户端原有位置和时区。
+全局请求位置在管理端「系统设置 → 上游配置 → 请求位置覆盖」配置，默认关闭，保留客户端原有位置和时区。
 开启后使用自定义位置，关闭时保留已保存的自定义值。「代理管理」设置的自定义位置优先于全局设置，
 全局关闭时代理自定义仍生效。保存后新请求生效，无需重启；关闭代理自定义后恢复全局继承。
 该设置统一 OpenAI Responses 的 Web Search 位置与带来源标记的环境日期、时区，不修改 epoch 时间戳、
@@ -342,12 +342,10 @@ OpenAI 主动额度重置卡及其消费结果由上游持有，不写入 Postgr
 ## 密码语义
 
 - `admin.default_password` 只在首次创建管理员时使用。
-- 已有管理员在「系统设置 → 管理员密码」修改登录密码，需要验证当前密码；修改后所有管理员会话失效，使用新密码重新登录。
+- 已有管理员在「系统设置 → 安全与访问 → 管理员密码」修改登录密码，需要验证当前密码；修改后所有管理员会话失效，使用新密码重新登录。
 - PostgreSQL 官方镜像只在空数据目录初始化时使用 `database.password`。
 - Redis 在每次容器创建时使用 `redis.password`。
 
-管理员会话绑定密码哈希指纹；从尚未包含此字段的版本升级后，管理员需要重新登录一次，密钥身份会话不受影响。
-
 已有 PostgreSQL 数据目录后，直接修改 `database.password` 不会修改数据库用户密码，只会导致
 应用无法连接。轮换时必须先在 PostgreSQL 中修改用户密码，再同步更新 `config.yaml`。Redis
 密码变更后需要用新配置重新创建 Redis 和应用容器，不需要删除 Redis 数据目录。

docs/api.md

@@ -267,7 +267,8 @@ OpenAI 明确返回 `server_is_overloaded`、`slow_down` 或模型容量不足
 尚未交付输出的前提下，先做最多 3 次同账号指数退避，再通过现有调度换号。默认间隔从 500ms 开始，
 上游 `Retry-After` 参与退避计算，单次等待不超过 8 秒；重试同时受请求总尝试次数和截止时间约束。
 `server_is_overloaded`、`slow_down` 等可计分的结构化错误按已发送的失败尝试计入 Smart 账号
-健康分。失败率使用账号级平滑与时间衰减，影响后续普通选路，已有可用账号的
+健康分。已确认容量拒绝的平滑权重为 0.4，其他可计分失败与成功样本保持 0.2。
+失败率使用账号级平滑与时间衰减，影响后续普通选路，已有可用账号的
 会话亲和仍优先。容量不足不触发 Provider 全局熔断，也不作为账号额度耗尽；启用账号自动冻结时，
 达到容量失败阈值会另外写入临时冷却。
 客户端错误兼容由 API 编码出口统一处理：最终交付的 `server_is_overloaded`、`slow_down` 错误码
@@ -857,7 +858,13 @@ PostgreSQL 或 Redis。
 分组是 Provider-neutral 的账号集合；一个组可包含任意 Provider 账号，一个账号也可属于多个组。
 分组详情和列表返回 `disableFast`，创建时省略默认为 `false`，更新时省略或 `null` 保留现值。
 Client Key 绑定的任一分组开启此限制（包括已禁用分组）时，该 Key 的 OpenAI Responses 请求关闭 Fast；
-未绑定分组的 Key 只受全局限制，不按最终所选账号的分组判断。
+未绑定分组的 Key 不限制 Fast，不按最终所选账号的分组判断。
+
+关闭 Fast 只将顶层 `service_tier` 的 `priority`（含 `fast` 别名）改为显式 `default`，继续处理请求；
+不改变 `flex`、`ultrafast`、缺失值、默认档、嵌套字段或其他 Provider。
+HTTP 和每个 WebSocket `response.create` 均使用请求开始时的分组策略，同一请求重试保持该策略；
+HTTP 请求头及新建 WS 的握手提示按当时的最终出站档位构造；复用 WS 时不重发握手头，
+每个 `response.create` 仍独立应用档位策略，请求档位统计与本地费用估算使用各帧的最终出站档位。
 
 | 方法 | 路由 | 主要 query/body | 说明 |
 | --- | --- | --- | --- |
@@ -971,7 +978,6 @@ HTTP 返回 `429`，`error.code` 为 `key_daily_budget_exceeded` 或 `key_weekly
 
 ```text
 openaiClientProfile
-disableFast
 requestLocationEnabled
 requestLocation
 modelMappings
@@ -998,13 +1004,6 @@ accountAutoFreezeProbeModel
 accountAutoFreezeAdaptiveConcurrency
 ```
 
-`disableFast` 默认 `false`，更新时省略或 `null` 保留现值。全局开启时，所有 Key 的 OpenAI Responses 请求关闭 Fast；
-全局关闭时仍应用 Key 绑定分组的限制。关闭 Fast 只将顶层 `service_tier` 的 `priority`（含 `fast` 别名）
-改为显式 `default`，继续处理请求；不改变 `flex`、`ultrafast`、缺失值、默认档、嵌套字段或其他 Provider。
-HTTP 和每个 WebSocket `response.create` 均使用请求开始时的配置，同一请求重试保持该配置；
-HTTP 请求头及新建 WS 的握手提示按当时的最终出站档位构造；复用 WS 时不重发握手头，
-每个 `response.create` 仍独立应用档位策略，请求档位统计与本地费用估算使用各帧的最终出站档位。
-
 `requestLocationEnabled` 是必填布尔值，默认 `false`：关闭时不覆盖客户端原有位置和时区；开启时使用已保存的
 `requestLocation`。关闭不会清空自定义值，代理自定义位置仍优先。
 `requestLocation` 是必填的完整对象 `{ country, region, city, timezone }`，不接受 `null`；初始保存值为
@@ -1038,7 +1037,7 @@ HTTP 请求头及新建 WS 的握手提示按当时的最终出站档位构造
 | `GET` | `/api/admin/settings/pricing` | 返回 `{ defaults, synced, overrides, syncedAt }` |
 | `POST` | `/api/admin/settings/pricing/update` | `{ provider, models, change }`，成功返回 `{ saved: true }` |
 | `POST` | `/api/admin/settings/pricing/sync/preview` | 无 body；返回 `{ prices, skipped }`，不写入配置 |
-| `POST` | `/api/admin/settings/pricing/sync` | 原样提交确认的 `{ prices, skipped }`；成功返回 `{ saved: true }` |
+| `POST` | `/api/admin/settings/pricing/sync` | `{ preview: { prices, skipped }, models: { openai: ["gpt-5.4"] } }`；成功返回 `{ saved: true }` |
 
 价目使用 `Provider → 精确上游模型 ID → { multiplierBps, bands }` 的映射。优先级为人工覆盖、已同步价目、
 内置价目；按档位合并，不从客户端模型别名或响应模型猜测价格。`syncedAt` 为 ISO 时间或 `null`。
@@ -1060,12 +1059,17 @@ OpenAI 长上下文为输入超过 272000 Token，xAI 为输入达到 200000 Tok
   配置，未提供档位重新继承来源；内置和同步均未登记的模型必须包含 `standard`。
 - `{ "action": "multiplier", "multiplierBps": 20000 }`：设置目标倍率，保留已有人工单价；重复提交不连续相乘。
 - `{ "action": "reset" }`：移除人工单价与倍率，恢复同步价或内置价；仅有人工价格的模型恢复为未配置。
+- `{ "action": "delete" }`：删除非内置模型的同步价目与人工配置；批量包含任何内置模型时整批返回 400，
+  即使该内置模型已有人工覆盖也不能删除。删除不影响历史账单，之后可重新添加或选中同步导入。
 
 一批更新原子提交并写审计，不覆盖未选中的模型。未知字段、错误类型和非法价格字符串等 JSON 合同错误
 返回 422；Provider、模型 ID、批量数量、倍率上限和不支持的档位等业务校验错误返回 400。
 models.dev 同步只导入可表示为当前文本 Token 计价的 OpenAI/xAI 模型；不完整价格、其他输出模态及
 不匹配的上下文梯度在 `skipped` 中返回 `provider/model`。确认会重新抓取价目；若与预览不同返回 400，
-需重新预览。来源不可用返回 502，已有价目保持不变。同步只更新来源层，始终保留人工单价与倍率。
+需重新预览。来源不可用返回 502，已有价目保持不变。`preview` 必须原样提交，`models` 按 Provider
+指定 1～10000 个待同步模型；空选择或未登记的模型返回 400。同步只更新选中模型的来源层，未选中模型
+及所有人工单价与倍率保持不变。选中的已同步模型若不再出现在来源价目中，则移除其来源层，恢复内置价格；
+没有内置价目的模型变为未配置。
 
 ### OpenAI 上游客户端身份
 

docs/architecture.md

@@ -133,8 +133,8 @@ service；客户端原生对象保存在有界进程缓存中，与套餐 Redis
 `RuntimeSnapshot → RoutingPlan → AttemptContext` 冻结传递，关闭时保留客户端原有字段；
 Provider 在选定账号后应用代理位置覆盖。请求期间不额外查询全局设置，配置发布不改变已开始请求的全局值。
 
-Fast 限制由同一快照链路冻结：全局开关与 Client Key 所有绑定分组的开关取逻辑或；
-禁用分组仍贡献限制，无分组 Key 只受全局开关影响，不按所选账号的分组重新解释。
+Fast 限制由同一快照链路冻结：Client Key 所有绑定分组的开关取逻辑或；
+禁用分组仍贡献限制，无分组 Key 不限制 Fast，不按所选账号的分组重新解释。
 OpenAI Provider 在独立编码请求上、生成上游头与观测前统一应用顶层档位覆盖，HTTP、WS 与重试共用。
 该策略不改变选号、会话亲和或共享原始请求；每个新 WS 请求重新取得当前策略。
 WS 路由提示属于握手，连接复用时不重发；档位变化不重建连接或打断 `previous_response_id` 续写，

docs/plans/upstream-client-profiles.md

@@ -1,266 +0,0 @@
-# 上游客户端身份配置实施计划
-
-状态：实现与本地验收完成，真实上游账号验收尚未覆盖。当前实现位于 `feat/upstream-client-profiles` 独立工作区。
-代码核对基线：`3328a435`。
-
-## 1. 目标与已确定的规则
-
-为 OpenAI 上游请求提供 macOS、Linux、Windows 与 Desktop、CLI 组合的六套客户端身份预设，
-支持自定义，并允许管理员在 Client Key 中单独选择，覆盖通用设置。
-
-配置只保留两层：
-
-```text
-Key 有独立配置 → 使用 Key 配置
-Key 跟随通用设置 → 使用通用设置
-```
-
-- 新安装的通用设置默认使用 **macOS · Desktop · 自动最新**，管理员可以修改。
-- 新建及升级前已有的 Key 默认「跟随通用设置」。
-- 内置默认值只用于初始化通用设置，不参与运行时配置优先级。
-- 修改通用设置后，跟随它的 Key 在新请求中自动生效；独立配置的 Key 保持自己的选择。
-- Key 切回「跟随通用设置」时清除独立覆盖，不保留参与运行时合并的隐藏字段。
-- 独立覆盖以完整配置为单位，不逐字段混合 Key 与通用设置，避免组合出不一致的身份。
-
-本文中的客户端身份指网关发给上游的 User-Agent、originator、平台和版本等信息。
-该功能不改变接入客户端的最低版本门禁、客户端安装方式或网关宿主操作系统。
-
-## 2. 当前实现与可复用入口
-
-| 现有能力 | 代码入口 | 本次需要补齐的部分 |
-| --- | --- | --- |
-| 启动时的 Desktop 身份配置，默认 macOS | [OpenAI 配置](../../backend/crates/providers/openai/src/config.rs) | 将用户选择纳入持久化通用设置，支持客户端类型与平台选择 |
-| 共享请求画像、Desktop 发布检查及版本元组更新 | [请求画像](../../backend/crates/providers/openai/src/transport/profile/mod.rs) | 按类型与平台解析版本，支持固定版本以及每请求独立身份 |
-| 模型、账号及 Desktop 专属请求的头部组装 | [请求头](../../backend/crates/providers/openai/src/transport/headers.rs) | 按调用场景消费解析后的身份，保留不同接口的协议边界 |
-| 通用设置管理 | [管理模型](../../backend/crates/gateway-admin/src/model/settings.rs)、[持久化](../../backend/crates/gateway-store/src/postgres/runtime_settings.rs) | 保存全局选择，复用事务、审计及配置发布 |
-| Key 管理 | [管理模型](../../backend/crates/gateway-admin/src/model/client_keys.rs)、[管理用例](../../backend/crates/gateway-admin/src/use_case/client_keys.rs) | 保存可空的独立覆盖，并提供实际生效配置的投影 |
-| 请求策略冻结与发布 | [运行快照](../../backend/crates/gateway-core/src/routing/snapshot.rs)、[快照发布](../../backend/crates/gateway-core/src/runtime/mod.rs) | 携带配置选择，保持同一请求及重试的一致性 |
-| 上游连接池已有身份区分与精确续写；模型目录已有缓存隔离 | [WebSocket](../../backend/crates/providers/openai/src/transport/websocket/)、[账号目录](../../backend/crates/providers/openai/src/credential/catalog.rs) | 复用现有机制，核对新增身份字段是否完整进入隔离维度 |
-| 设置页及首页身份展示 | [设置页](../../frontend/src/views/settings/)、[身份卡片](../../frontend/src/views/dashboard/components/WireProfileCard.vue) | 增加配置入口，明确全局展示与 Key 实际生效值的区别 |
-
-目前画像类型包含 Desktop 专属字段，自动更新通过同一 Desktop 制品取得应用版本、Core 版本及构建号。
-因此，六套配置需要按官方身份语义扩展，不能仅替换现有 UA 的操作系统名称。
-
-## 3. 预设、自定义与版本策略
-
-### 3.1 六套预设
-
-| 平台 | Desktop | CLI |
-| --- | --- | --- |
-| macOS | macOS Desktop | macOS CLI |
-| Linux | Linux Desktop | Linux CLI |
-| Windows | Windows Desktop | Windows CLI |
-
-平台与客户端类型决定预设；系统版本、CPU 架构及终端信息是预设参数，不再展开成额外的顶级选项。
-预设值由后端统一提供，前端不维护另一份默认值和 UA 拼接规则。
-
-实施前逐项核对官方源码或官方制品，记录版本、提交或制品标识、请求格式和更新来源。
-尤其需要核实 Linux Desktop 的可用制品及身份依据，不能因为有六个组合就宣称六套均已获得官方验证。
-没有可靠官方依据的组合保留自定义配置入口，标明核验状态；不伪造官方发布版本或自动更新来源。
-
-### 3.2 自定义
-
-通用设置和 Key 独立配置共用一套编辑模型：选择预设后，可以调整客户端标识、系统版本、
-架构、终端及适用的版本字段，并预览最终请求身份。
-
-- 首期将自定义内容保存在对应通用设置或 Key 中；命名模板库、共享模板引用及模板管理不属于本次范围。
-- 后端按客户端类型生成 UA 与相关身份头；不增加任意请求头透传或任意发布源 URL 配置。
-- 校验必填项、字段长度、HTTP 头合法性、控制字符及客户端类型对应的版本格式。
-- Desktop 与 CLI 使用各自适用的字段；CLI 不携带由 Desktop 默认值遗留的应用版本后缀。
-- 自定义值与「官方核验」状态分别展示，用户填写的值不能自行获得官方核验标记。
-- 认证信息、账号 ID、Cookie、请求位置及数据驻留仍由现有职责管理，不纳入身份自定义。
-
-### 3.3 自动最新与固定版本
-
-| 模式 | 行为 |
-| --- | --- |
-| 自动最新（默认） | 跟随该客户端、平台及适用架构对应的最新已核验稳定发布；后台更新，不在业务请求中访问发布源 |
-| 固定版本 | 使用保存的版本；后台发布检查不能覆盖固定值 |
-
-自动最新只更新版本资料，不覆盖用户的平台、架构、终端等选择。手动修改版本时转为固定版本，
-避免保存后又被后台任务覆盖。恢复自动最新后，重新采用对应发布源的有效版本。
-
-- Desktop 应用版本、内嵌 Core 版本和构建号作为同一制品的完整元组更新；不拼接独立 CLI 的最新版本。
-- CLI 使用对应官方 CLI 发布信息，不借用 Desktop 发布号。
-- 不假设不同平台的 Desktop 同步发布或共享构建号。
-- 更新失败保留同一预设上次成功的版本，并展示检查失败及检查时间；不得切到其他平台或客户端。
-- 首次离线启动可使用同一预设随版本附带的已核验版本资料，并标明未完成最新检查。
-  这些资料属于版本解析器的种子数据，不是第三层用户配置。
-- 没有有效发布源或启动资料的组合不能进入可用的「自动最新」状态；提供原因与固定自定义入口。
-- Redis 中的发布缓存至少区分客户端类型、平台和影响制品选择的架构，不再仅凭 Provider 区分。
-  缓存失效和重启后的恢复也不得覆盖用户的固定配置。
-
-## 4. 配置合同与职责
-
-### 4.1 持久化与管理 API
-
-拟增加的概念字段如下，最终名称在实施时与现有 API 风格统一：
-
-| 位置 | 字段 | 语义 |
-| --- | --- | --- |
-| 通用设置 | `openaiClientProfile` | 必须存在的完整配置选择，包含预设或自定义参数、版本模式 |
-| Client Key | `openaiClientProfileOverride` | `null` 表示跟随通用设置；对象表示完整独立配置 |
-| 只读投影 | `effectiveOpenaiClientProfile` | 返回当前解析结果、配置来源、版本来源及检查状态，不作为另一份可写配置 |
-
-新建 Key 省略覆盖字段等同于 `null`；更新 Key 省略字段保留已有覆盖，显式 `null` 才清除。
-旧调用方更新通用设置时省略新增字段，应保留现值，不能重置管理员已保存的选择。
-通用设置不接受以 `null` 制造第三层回退的语义。
-
-配置由 PostgreSQL 保存。管理写入沿用现有事务、脱敏审计、revision 和快照发布流程。
-官方发布结果继续作为 Provider 的派生缓存，不将每次版本检查写成用户配置或推进全局配置 revision。
-
-职责依照 [系统架构](../architecture.md)：
-
-- Admin 负责全局与 Key 的管理规则和继承选择，复用同一解析入口生成生效投影。
-- Core 负责请求策略的携带与冻结，不依赖具体 Provider，也不解释 UA、官方制品或请求头。
-- OpenAI Provider 负责预设、官方版本解析、字段语义校验及上游头部生成，通过现有端口风格提供能力。
-- Store 实现配置、迁移与缓存存取，不解释客户端身份的业务规则。
-- API 负责 wire 校验及投影，前端负责编辑与展示，不重复推导最终身份。
-
-### 4.2 生效范围
-
-本次配置作用于 OpenAI Provider。Key 同时授权其他 Provider 时，不把 OpenAI 身份套用到 xAI。
-
-由 Client Key 发起的 OpenAI 模型请求及其相关目录读取，消费该 Key 的有效身份；
-HTTP/SSE、WebSocket 以及 Images、Search 等入口按实际调用链逐一核对覆盖。
-OAuth 与 API Key 上游认证模式继续共用适用的身份规则，认证字段仍按各自协议生成。
-
-后台令牌刷新、额度查询、账号管理及 Desktop 专属接口没有 Client Key 上下文，不能被某个 Key 的选择污染。
-实施时按官方接口要求划分共用身份与 Desktop 专属身份，保留必要的独立请求格式；
-全局选为 CLI 时也不能让 Desktop 专属接口意外携带缺失或错误的应用版本。
-
-### 4.3 请求一致性与资源隔离
-
-本节以接入现有机制为主。当前鉴权已持有 `RuntimeSnapshot`，重试通过原路由计划传递冻结策略；
-WebSocket 连接池已将 `originator`、`user-agent` 等纳入连接身份，且已有
-[画像更新后保留精确续写所属连接的测试](../../backend/crates/providers/openai/tests/transport/websocket_pool.rs)。
-实施沿用这些快照、重试和续写机制，补齐新增配置的数据传递与必要隔离维度。
-
-- 将通用与 Key 的身份选择接入已有配置快照，沿现有请求上下文传入 Provider，
-  同一逻辑请求的重试与换号继续使用冻结的选择。
-- 自动发布版本目前属于 Provider 的共享运行时画像，`response_headers` 在组装头部时读取
-  `self.profile.snapshot()`，尚未由 `RuntimeSnapshot` 冻结。
-  为满足跨重试也使用同一解析版本的目标，在已有请求生命周期内保存首次解析结果并复用；
-  这部分与用户配置的快照继承分别验证，不改动重试决策和状态机。
-- 禁止通过临时修改共享 `CodexWireProfileState` 实现 Key 覆盖；并发 Key 使用独立的不可变请求身份。
-- 同一入站 WebSocket 的每个 `response.create` 按现有规则获取新策略。
-  普通连接获取复用现有身份匹配规则；精确续写继续使用持有响应状态的原连接，
-  保留已有的画像更新例外，不能因版本刷新强制断开或重建这类连接。
-- 检查新增字段是否已被现有连接身份和模型目录缓存键覆盖，仅补齐影响协议结果的缺失维度。
-  Key 授权边界沿用现有机制；身份切换继续遵循原有续写和重放限制，避免上下文丢失或越界重放。
-- 「当前生效」投影显示解析结果；请求观测使用本次冻结的摘要，不能事后用最新全局值替代。
-
-## 5. 管理端交互
-
-### 通用设置
-
-新增「上游客户端身份」配置区，包含平台、客户端类型、预设/自定义、版本策略以及相应字段。
-默认展示 macOS Desktop 自动最新；提供最终 UA、相关版本、发布检查状态及字段错误提示。
-保存后通过现有设置发布机制对新请求生效，无需重启。
-
-### Key 创建与编辑
-
-默认选择「跟随通用设置」，同时展示当前通用配置摘要。
-选择「独立配置」后使用与通用设置相同的编辑组件；首次切换可用当前通用值填充表单草稿，
-只有保存后才形成独立配置。摘要明确标出「通用设置」或「Key 独立配置」。
-
-### 现有展示
-
-首页身份卡片标明展示的是通用配置，不将其描述成所有 Key 的唯一身份。
-Key 详情展示该 Key 当前有效身份；复用后端投影，不在浏览器拼装 UA 或解析版本。
-复用现有基础组件及 [管理端主题](../theme.md)，提供保存失败、版本检查失败、无可用发布源等状态。
-
-## 6. 升级与迁移
-
-- 按 [迁移规则](../../backend/migrations/README.md) 新增编号迁移，更新冻结清单，不修改历史迁移。
-- 新安装在初始化阶段写入通用默认配置；存量 Key 的覆盖初始化为 `null`。
-- 升级需识别既有 `openai.wire_profile` 自定义值，首次导入通用配置时保留有效的环境与身份选择。
-  标准配置映射到 macOS Desktop 自动最新；显式自定义版本映射到固定自定义，并在升级说明交代。
-- SQL 迁移不负责读取部署文件；必要的一次性导入由启动组装流程经正式管理/存储端口完成，
-  在设置尚未初始化时原子写入。初始化标记避免重启或更新重新覆盖数据库中的设置。
-- 实施前核实配置加载器能否区分显式自定义与默认填充值，必要时在加载阶段保留该来源信息。
-- 初始化完成后，相关用户选择以数据库为唯一权威；清理旧配置的运行时覆盖路径并同步配置样例。
-  其他仍有效的 Provider 配置字段继续归原 owner 管理。
-- 备份与恢复包含通用配置和 Key 覆盖；发布缓存仍可重建，不成为配置恢复的前置依赖。
-
-## 7. 实施顺序
-
-1. **核对官方依据与接口范围。** 完成六个组合的来源矩阵，明确可自动更新的组合、
-   Desktop 专属接口、CLI 格式及旧配置来源识别方式，记录无法核验的项目。
-2. **定义合同并打通持久化。** 确定完整配置对象、继承语义及只读投影；实现新增迁移、
-   一次性初始化、设置与 Key 管理 API，复用审计和快照发布。
-3. **扩展 Provider 解析。** 复用现有 Desktop 版本核验，补充 CLI 及平台来源、固定版本、
-   自定义校验和分维度缓存；确保官方资料与用户配置分别存储。
-4. **接入实际请求链路。** 复用现有快照与请求上下文携带有效选择，补齐解析版本的请求级复用，
-   覆盖相关传输与目录入口；复用连接池身份匹配与精确续写规则，仅补齐必要的隔离维度。
-5. **实现设置与 Key 界面。** 复用一套编辑组件和服务端投影，调整首页身份卡片含义，
-   完成浏览器交互验收。
-6. **验证并同步正式文档。** 完成下列矩阵；在行为可用后就地更新 API、架构、部署和用户入口的必要说明，
-   不提前将本计划中的拟定合同写成已发布能力。
-
-## 8. 验收条件
-
-| 范围 | 必须覆盖的场景与结果 |
-| --- | --- |
-| 默认与继承 | 新安装采用约定默认；新旧 Key 跟随；通用修改只影响跟随者；独立配置不受影响；显式清除后恢复继承 |
-| API 更新语义 | 省略覆盖字段保留原值，显式 `null` 清除；旧设置调用不重置新配置；非法配置不留下部分提交 |
-| 预设与自定义 | 六个组合均有明确支持状态；已核验预设与对应官方证据一致；自定义可保存、重启恢复并生成合法头部 |
-| 版本解析 | 自动最新按来源更新；固定值不被刷新；Desktop 元组一致；检查失败保留同组合旧版本；无来源状态如实显示 |
-| 并发与冻结 | 两个不同配置的 Key 并发请求不串用身份；配置或发布刷新期间，已开始请求及其重试保持一致 |
-| 上游传输 | 在可用的真实测试环境检查 HTTP/SSE、WebSocket 及受影响接口发出的身份与结果；API Key 认证不混入 OAuth 字段 |
-| 缓存与续写 | 不同有效身份不误用目录结果或连接；配置切换后的新请求应用新身份；旧 continuation 无上下文丢失或越界重放 |
-| 管理与后台 | Key 覆盖不影响其他 Key、后台任务及 Desktop 专属操作；全局 CLI 与混合 Provider 授权场景可用 |
-| 升级与恢复 | 标准配置、自定义配置、重复启动、数据库备份恢复和发布缓存缺失均按约定处理；历史迁移校验不变 |
-| 界面 | 创建/编辑/清除覆盖、保存与加载失败、自动与固定切换、来源及核验状态正确；检查窄屏、深浅色和键盘操作 |
-
-后端补充能验证上述真实行为的用例，按 [贡献与审查](../../CONTRIBUTING.md#验证) 和
-[架构验收入口](../architecture.md#12-修改与验收) 执行适用的 Rustfmt、Clippy、测试、架构边界与迁移检查。
-数据库与 Redis 使用专用测试实例，因环境缺失跳过的测试明确记录为缺口。
-
-前端执行 `pnpm --dir frontend format:check`、`pnpm --dir frontend build`，并提供实际浏览器运行证据；
-不为本功能引入项目当前不维护的独立前端测试体系。所有变更执行 `git diff --check`。
-构建、单元测试或模拟上游通过不能替代真实协议验收；测试账号和上游环境不足时列明未覆盖组合。
-
-## 9. 已核对的官方依据与实现边界
-
-- CLI UA 与 originator：官方 `openai/codex` 提交 `4d205c7a4dc36b719679a0356a45b23133732265`，
-  `codex-rs/login/src/auth/default_client.rs`。CLI 不添加 Desktop 应用后缀。
-- CLI 版本：`https://registry.npmjs.org/@openai%2Fcodex/latest`，2026-09-18 返回 `0.155.0`，
-  darwin/linux/win32 × arm64/x64 的依赖版本均配套。实现验证包名、稳定 SemVer 和六个依赖，逐制品缓存。
-- macOS Desktop：沿用官方 appcast 与同一 ZIP 的内嵌 Core 核验路径，支持 arm64。
-- Windows Desktop：分别读取官方 `ChatGPT-x64.msix`、`ChatGPT-arm64.msix`，从 `app.asar/package.json`
-  获取应用版本和构建号，从同包 `codex.exe` 核验架构和 Core 版本；MSIX 清单版本不作为应用版本。
-- Linux Desktop：依据[官方安装说明](https://learn.chatgpt.com/docs/linux/linux-app)，分别读取 amd64/arm64 DEB，
-  按 XZ 索引跳读 tar，核验同包 `app.asar/package.json` 与 ELF Core。Windows/Linux 的 Range 请求锁定 ETag，
-  未变更时只检查一个字节，变更后完整核验版本元组；独立缓存、检查失败保留已核验版本。
-
-身份预览采用独立只读接口，由后端解析全局或所提交的完整覆盖；Key 列表保存并返回覆盖配置，
-编辑器按预览结果显示当前生效身份，不在前端生成 UA。
-
-## 10. 本地验收记录
-
-2026-09-18，独立 PostgreSQL 18、Redis 8 与 `127.0.0.1:28080` 测试实例，使用合成账号与 Key。
-
-- `RUST_MIN_STACK=16777216 cargo +1.97.0 test --manifest-path backend/Cargo.toml --test main --locked`
-  全部 9 个测试套件通过，合计 2420 项，PostgreSQL/Redis 使用上述专用实例。
-  覆盖配置继承、首次导入、旧调用保留字段、清除覆盖、跨重试版本冻结、并发请求身份隔离、制品缓存隔离，
-  以及 Windows/Linux 归档解析、架构不符、损坏/截断、ETag 变化、XZ 跨块跳读和发布回退。
-- `cargo +1.97.0 fmt --all --manifest-path backend/Cargo.toml -- --check`、严格 Clippy（所有 target/feature）通过。
-- `pnpm --dir frontend format:check`、`pnpm --dir frontend build`、`git diff --check` 通过。
-- `backend/migrations/.frozen-sha256` 全部校验通过，历史迁移未修改。
-- Chrome 实测通用设置切换并保存 Linux CLI、创建默认继承的 Key、保存 Windows CLI 固定版本覆盖、
-  重载回显与清除覆盖后恢复继承；检查 1920 px 与 390 px、深浅色、Key 弹窗和键盘关闭下拉框。
-  新增表单宽度与相邻设置对齐，文案仅保留短标签与必要状态，浏览器无错误或警告。
-- Chrome 实测 Windows/Linux Desktop 自动模式与各自 UA，Windows 保存后重载回显；验收后恢复原 MacOS Desktop 通用配置。
-  下拉显示 `MacOS`，内部枚举仍为 `macos`，UA 的系统类型仍为 `Mac OS`。
-- 官方发布检查实际成功：macOS Desktop `26.915.31029` / Core `0.155.0-alpha.9`，CLI `0.155.0`。
-  Windows 两种架构为 Desktop `26.908.70816` / Core `0.154.0-alpha.6.2` / Build `9275`；
-  Linux 两种架构为 Desktop `26.915.31029` / Core `0.155.0-alpha.9` / Build `9771`。
-  四个官方包首次核验耗时约 17～30 秒，重复检查约 67～74 毫秒，未重新扫描制品。
-- `cargo audit --file backend/Cargo.lock` 未通过：已有固定依赖 `rustls 0.23.36` 命中
-  `RUSTSEC-2026-0285`，另有原有 `chacha20`、`wnaf` 撤回警告；新增归档依赖未报告漏洞。
-  本次未改变受官方 TLS 指纹约束的依赖版本。
-
-验证边界：没有真实上游账号，真实认证下的 HTTP/SSE、WebSocket、Images/Search 未进行端到端验收；
-本地传输测试不能替代该验收。Windows/Linux Desktop 的 x86_64、arm64 官方制品解析和未变化复用均已实测，真实账号请求仍未覆盖。
-完整备份恢复、生产部署环境及发布源故障下的界面交互尚未实测。

frontend/src/api/modules/pricing.ts

@@ -28,12 +28,18 @@ export type PricingChange
   = | { action: 'replace', pricing: ModelPricing }
     | { action: 'multiplier', multiplierBps: number }
     | { action: 'reset' }
+    | { action: 'delete' }
 
 export interface PricingSyncPreview {
   prices: PricingMap
   skipped: string[]
 }
 
+export interface PricingSyncRequest {
+  preview: PricingSyncPreview
+  models: Record<string, string[]>
+}
+
 interface UpdatePricingParam {
   provider: string
   models: string[]
@@ -68,7 +74,7 @@ export function previewPricingSync() {
   })
 }
 
-export function syncPricing(data: PricingSyncPreview) {
+export function syncPricing(data: PricingSyncRequest) {
   return request<PricingMutationResponse>({
     url: '/api/admin/settings/pricing/sync',
     method: 'POST',

frontend/src/api/modules/settings.ts

@@ -8,7 +8,6 @@ export type RotationStrategy = 'smart' | 'quota_reset_priority' | 'round_robin'
 export interface RuntimeSettings {
   openaiClientProfile: ClientProfileSelection
 
-  disableFast: boolean
   requestLocationEnabled: boolean
   requestLocation: RequestLocation
   modelMappings: Record<string, string>

frontend/src/components/AccountGroupCheckboxGrid.vue

@@ -47,6 +47,6 @@ function updateGroup(groupId: string, selected: boolean) {
     </div>
   </div>
   <p v-else class="m-0 rounded-cp bg-cp-fill-quaternary px-3.5 py-3 text-cp-sm font-emphasis text-cp-text-secondary">
-    {{ loading ? '正在加载分组...' : '暂无分组，请先在分组管理中创建。' }}
+    {{ loading ? '正在加载分组...' : '暂无分组，请先在分组管理中创建' }}
   </p>
 </template>

frontend/src/components/base/BaseMarkdown/index.vue

@@ -0,0 +1,33 @@
+<script setup lang="ts">
+import { computed, onMounted, shallowRef, useTemplateRef } from 'vue'
+import { renderMarkdown } from './markdown'
+import markdownStyles from './markdown.css?inline'
+
+const props = defineProps<{
+  source?: string | null
+}>()
+
+const host = useTemplateRef<HTMLElement>('host')
+const contentRoot = shallowRef<HTMLElement | null>(null)
+const html = computed(() => renderMarkdown(props.source))
+
+onMounted(() => {
+  if (!host.value)
+    return
+
+  const shadowRoot = host.value.attachShadow({ mode: 'open' })
+  const style = document.createElement('style')
+  const content = document.createElement('div')
+  style.textContent = markdownStyles
+  shadowRoot.append(style, content)
+  contentRoot.value = content
+})
+</script>
+
+<template>
+  <div ref="host" class="min-w-0">
+    <Teleport v-if="contentRoot" :to="contentRoot">
+      <div class="markdown-body" v-html="html" />
+    </Teleport>
+  </div>
+</template>

frontend/src/components/base/BaseMarkdown/markdown.css

@@ -0,0 +1,102 @@
+.markdown-body {
+  color: var(--cp-color-text);
+  font-family: var(--font-sans);
+  font-size: var(--cp-font-size-sm);
+  font-weight: var(--font-weight-emphasis);
+  line-height: 1.65;
+  overflow-wrap: anywhere;
+}
+
+.markdown-body * {
+  box-sizing: border-box;
+  max-width: 100%;
+}
+
+.markdown-body > :first-child {
+  margin-top: 0;
+}
+
+.markdown-body > :last-child {
+  margin-bottom: 0;
+}
+
+.markdown-body h1,
+.markdown-body h2,
+.markdown-body h3,
+.markdown-body h4,
+.markdown-body h5,
+.markdown-body h6 {
+  margin: 12px 0 6px;
+  color: var(--cp-color-text);
+  font-size: inherit;
+  font-weight: var(--font-weight-heavy);
+  line-height: 1.4;
+}
+
+.markdown-body p {
+  margin: 0 0 8px;
+}
+
+.markdown-body ul,
+.markdown-body ol {
+  margin: 0 0 8px;
+  padding-left: 1.5em;
+}
+
+.markdown-body ul {
+  list-style-type: disc;
+}
+
+.markdown-body ol {
+  list-style-type: decimal;
+}
+
+.markdown-body li {
+  margin: 3px 0;
+}
+
+.markdown-body a {
+  color: var(--cp-color-link);
+  font-weight: var(--font-weight-heavy);
+  text-decoration: none;
+}
+
+.markdown-body a:hover {
+  color: var(--cp-color-link-hover);
+  text-decoration: underline;
+}
+
+.markdown-body a:focus-visible {
+  border-radius: var(--cp-border-radius-sm);
+  outline: 2px solid var(--cp-control-outline);
+  outline-offset: 2px;
+}
+
+.markdown-body code {
+  border-radius: var(--cp-border-radius-sm);
+  background: var(--cp-color-fill-tertiary);
+  color: var(--cp-color-text);
+  font-family: var(--font-mono);
+  font-size: var(--cp-font-size-xs);
+  padding: 1px 5px;
+}
+
+.markdown-body pre {
+  margin: 8px 0;
+  overflow-x: auto;
+  border-radius: var(--cp-border-radius);
+  background: var(--cp-color-bg-container);
+  padding: 8px 10px;
+}
+
+.markdown-body pre code {
+  background: transparent;
+  padding: 0;
+}
+
+.markdown-body blockquote {
+  margin: 8px 0;
+  border-left: 3px solid var(--cp-color-split);
+  color: var(--cp-color-text-secondary);
+  padding-left: 10px;
+}

frontend/src/components/base/BaseMarkdown/markdown.ts

@@ -7,7 +7,7 @@ const marked = new Marked({
   gfm: true,
 })
 
-export function renderReleaseNotes(source?: string | null) {
+export function renderMarkdown(source?: string | null) {
   const value = source?.trim()
   if (!value)
     return ''

frontend/src/components/base/BaseModal/index.vue

@@ -102,10 +102,17 @@ function closeModal() {
   open.value = false
 }
 
-function focusableElements() {
-  return Array.from(panel.value?.querySelectorAll<HTMLElement>(focusableSelector) ?? []).filter(
-    element => !element.hidden,
-  )
+function focusableElements(root: ParentNode | null = panel.value): HTMLElement[] {
+  if (!root)
+    return []
+
+  const elements: HTMLElement[] = []
+  for (const element of root.children) {
+    if (element instanceof HTMLElement && element.matches(focusableSelector) && !element.hidden)
+      elements.push(element)
+    elements.push(...focusableElements(element.shadowRoot ?? element))
+  }
+  return elements
 }
 
 function handleKeydown(event: KeyboardEvent) {
@@ -125,17 +132,20 @@ function handleKeydown(event: KeyboardEvent) {
   }
   const first = focusable[0]
   const last = focusable[focusable.length - 1]
+  let activeElement = document.activeElement
+  while (activeElement?.shadowRoot?.activeElement)
+    activeElement = activeElement.shadowRoot.activeElement
   if (!panel.value?.contains(document.activeElement)) {
     event.preventDefault()
     if (event.shiftKey)
       last?.focus()
     else first?.focus()
   }
-  else if (event.shiftKey && document.activeElement === first) {
+  else if (event.shiftKey && activeElement === first) {
     event.preventDefault()
     last?.focus()
   }
-  else if (!event.shiftKey && document.activeElement === last) {
+  else if (!event.shiftKey && activeElement === last) {
     event.preventDefault()
     first?.focus()
   }

frontend/src/components/base/BaseTable/index.vue

@@ -55,7 +55,8 @@ const tableRef = useTemplateRef<HTMLTableElement>('table')
 const horizontalScrolled = shallowRef(false)
 const horizontalCanScrollRight = shallowRef(false)
 
-function measureHorizontalScroll() {
+function updateScrollLayout() {
+  scrollbarRef.value?.update()
   const wrap = scrollbarRef.value?.wrapRef
   if (!wrap) {
     horizontalScrolled.value = false
@@ -78,12 +79,12 @@ function handleTableScroll(payload: { scrollTop: number, scrollLeft: number }) {
 
 onMounted(async () => {
   await nextTick()
-  measureHorizontalScroll()
+  updateScrollLayout()
 })
-useResizeObserver(() => [scrollbarRef.value?.wrapRef, tableRef.value].filter(Boolean), measureHorizontalScroll)
+useResizeObserver(() => [scrollbarRef.value?.wrapRef, tableRef.value].filter(Boolean), updateScrollLayout)
 watch([() => displayRows.value.length, () => props.columns], async () => {
   await nextTick()
-  measureHorizontalScroll()
+  updateScrollLayout()
 })
 
 const headerRowClass = computed(() => [

frontend/src/components/client-profile/ClientProfileEditor.vue

@@ -7,10 +7,12 @@ import BaseFormItem from '@/components/base/BaseForm/FormItem.vue'
 import BaseInput from '@/components/base/BaseInput.vue'
 import BaseSegmented from '@/components/base/BaseSegmented.vue'
 import BaseSelect from '@/components/base/BaseSelect.vue'
+import BaseSkeleton from '@/components/base/BaseSkeleton.vue'
 import { errorMessage } from '@/utils/async'
 import { formatDateTime } from '@/utils/date'
 
-withDefaults(defineProps<{ disabled?: boolean, allowInherit?: boolean }>(), {
+const props = withDefaults(defineProps<{ active?: boolean, disabled?: boolean, allowInherit?: boolean }>(), {
+  active: true,
   disabled: false,
   allowInherit: false,
 })
@@ -22,7 +24,6 @@ const loading = shallowRef(true)
 const loadError = shallowRef('')
 const previewError = shallowRef('')
 const previewing = shallowRef(false)
-const custom = shallowRef(false)
 const platforms = { macos: 'MacOS', linux: 'Linux', windows: 'Windows' }
 const presetOptions = computed(() => presets.value.map(({ configuration }) => ({
   value: `${configuration.platform}-${configuration.client}`,
@@ -46,10 +47,8 @@ const selectedPreset = computed({
   get: () => model.value ? `${model.value.platform}-${model.value.client}` : '',
   set: (value: string) => {
     const preset = presets.value.find(({ configuration }) => `${configuration.platform}-${configuration.client}` === value)
-    if (preset) {
+    if (preset)
       model.value = { ...preset.configuration, versionMode: preset.automaticAvailable ? 'latest' : 'fixed' }
-      custom.value = false
-    }
   },
 })
 const versionMode = computed({
@@ -95,12 +94,12 @@ async function load() {
   }
 }
 
-watch(model, (configuration, _, onCleanup) => {
+watch([model, () => props.active], ([configuration, active], _, onCleanup) => {
   let cancelled = false
   preview.value = undefined
   previewError.value = ''
-  previewing.value = !needsVersionInput.value
-  if (needsVersionInput.value)
+  previewing.value = active && !needsVersionInput.value
+  if (!active || needsVersionInput.value)
     return
   const timer = setTimeout(async () => {
     try {
@@ -141,10 +140,10 @@ onMounted(load)
       <BaseSegmented
         v-if="allowInherit"
         v-model="profileSource"
-        label="上游身份来源"
+        label="客户端身份来源"
         class="justify-self-start"
         :options="[
-          { label: '通用设置', value: 'global' },
+          { label: '全局配置', value: 'global' },
           { label: '独立配置', value: 'independent' },
         ]"
         :disabled="disabled"
@@ -154,13 +153,13 @@ onMounted(load)
           <BaseFormItem label="客户端预设">
             <BaseSelect v-model="selectedPreset" class="w-full" :options="presetOptions" :disabled="disabled" />
           </BaseFormItem>
-          <BaseFormItem label="版本模式">
+          <BaseFormItem label="版本策略">
             <BaseSelect
               v-model="versionMode"
               class="w-full"
               :options="[
-                { label: '自动最新', value: 'latest', disabled: !currentPreset?.automaticAvailable },
-                { label: '固定版本', value: 'fixed' },
+                { label: '跟随最新版本', value: 'latest', disabled: !currentPreset?.automaticAvailable },
+                { label: '自定义版本', value: 'fixed' },
               ]"
               :disabled="disabled"
             />
@@ -170,7 +169,7 @@ onMounted(load)
           {{ currentPreset.reason }}
         </p>
         <div v-if="model.versionMode === 'fixed'" class="grid gap-4 sm:grid-cols-2">
-          <BaseFormItem label="Core 版本" required>
+          <BaseFormItem label="Codex Core 版本" required>
             <BaseInput :model-value="model.codexVersion ?? ''" :disabled="disabled" placeholder="例如 0.155.0" @update:model-value="updateField('codexVersion', $event)" />
           </BaseFormItem>
           <template v-if="model.client === 'desktop'">
@@ -181,38 +180,36 @@ onMounted(load)
               <BaseInput :model-value="model.desktopBuild ?? ''" :disabled="disabled" placeholder="填写该制品的构建号" @update:model-value="updateField('desktopBuild', $event)" />
             </BaseFormItem>
           </template>
+          <BaseFormItem v-for="field in customFields" :key="field.key" :label="field.label">
+            <BaseInput
+              :model-value="model[field.key] ?? ''"
+              :placeholder="currentPreset?.defaults[field.key] ?? ''"
+              :disabled="disabled"
+              @update:model-value="updateField(field.key, $event)"
+            />
+          </BaseFormItem>
         </div>
-        <details :open="custom" @toggle="custom = ($event.target as HTMLDetailsElement).open">
-          <summary class="cursor-pointer text-cp-sm text-cp-text-secondary">
-            自定义身份参数
-          </summary>
-          <div class="mt-4 grid gap-4 sm:grid-cols-2">
-            <BaseFormItem v-for="field in customFields" :key="field.key" :label="field.label">
-              <BaseInput
-                :model-value="model[field.key] ?? ''"
-                :placeholder="currentPreset?.defaults[field.key] ?? ''"
-                :disabled="disabled"
-                @update:model-value="updateField(field.key, $event)"
-              />
-            </BaseFormItem>
-          </div>
-        </details>
       </template>
     </template>
-    <div class="grid min-w-0 gap-2 rounded-cp bg-cp-fill-quaternary p-4" aria-live="polite">
+    <div class="grid min-h-20 min-w-0 content-start gap-2 rounded-cp bg-cp-fill-quaternary p-4" aria-live="polite" :aria-busy="previewing">
       <p v-if="needsVersionInput" class="m-0 text-cp-sm text-cp-text-tertiary">
         填写版本后预览
       </p>
-      <p v-else-if="previewing" class="m-0 text-cp-sm text-cp-text-tertiary">
-        正在解析身份…
-      </p>
+      <div v-else-if="previewing" class="grid gap-2" role="status" aria-label="正在解析客户端身份">
+        <div class="flex h-lh items-center text-cp-sm" aria-hidden="true">
+          <BaseSkeleton shape="text" class="w-4/5" />
+        </div>
+        <div class="flex h-lh items-center text-cp-xs" aria-hidden="true">
+          <BaseSkeleton shape="text" class="w-52 max-w-full" />
+        </div>
+      </div>
       <p v-else-if="previewError" role="alert" class="m-0 text-cp-sm text-cp-error">
         {{ previewError }}
       </p>
       <template v-else-if="preview">
         <code class="break-all text-cp-sm text-cp-text">{{ preview.userAgent }}</code>
         <p class="m-0 text-cp-xs text-cp-text-tertiary">
-          {{ preview.versionSource === 'custom' ? '固定版本' : '自动更新' }}
+          {{ preview.versionSource === 'custom' ? '自定义版本' : '自动更新' }}
           <template v-if="preview.versionSource === 'official'">
             · {{ preview.checkedAt ? `检查于 ${formatDateTime(preview.checkedAt)}` : '待检查' }}
           </template>

frontend/src/layout/components/SystemUpdateModal/index.vue

@@ -14,13 +14,13 @@ import { computed, nextTick, shallowRef, useTemplateRef, watch } from 'vue'
 import BaseButton from '@/components/base/BaseButton.vue'
 import BaseConfirmModal from '@/components/base/BaseConfirmModal.vue'
 import BaseEmpty from '@/components/base/BaseEmpty.vue'
+import BaseMarkdown from '@/components/base/BaseMarkdown/index.vue'
 import BaseModal from '@/components/base/BaseModal/index.vue'
 import BaseScrollbar from '@/components/base/BaseScrollbar.vue'
 import { toast } from '@/components/base/BaseToast'
 import { normalizeSystemVersion, useSystemUpdateStore } from '@/stores/modules/system-update'
 import { errorMessage } from '@/utils/async'
 import { formatTime } from '@/utils/date'
-import { renderReleaseNotes } from './markdown'
 import {
   resolveSystemUpdateLogClasses,
   resolveSystemUpdatePresentation,
@@ -77,7 +77,7 @@ const updateLogRows = computed(() =>
   })),
 )
 
-const renderedReleaseNotes = computed(() => renderReleaseNotes(updateInfo.value?.notes))
+const hasReleaseNotes = computed(() => Boolean(updateInfo.value?.notes?.trim()))
 
 const showUpdateProgress = computed(
   () => hasUpdate.value || updating.value || restarting.value || updateLogRows.value.length > 0,
@@ -256,7 +256,7 @@ watch(
       </section>
 
       <section
-        v-if="renderedReleaseNotes"
+        v-if="hasReleaseNotes"
         class="grid gap-2 rounded-cp-card bg-cp-fill-quaternary px-4 py-3.5"
       >
         <div class="flex items-center justify-between gap-3">
@@ -269,7 +269,7 @@ watch(
         </div>
         <BaseScrollbar class="-mx-4" max-height="160px">
           <div class="px-4">
-            <div class="release-notes" v-html="renderedReleaseNotes" />
+            <BaseMarkdown :source="updateInfo?.notes" />
           </div>
         </BaseScrollbar>
       </section>
@@ -401,89 +401,3 @@ watch(
     </div>
   </BaseConfirmModal>
 </template>
-
-<style scoped>
-.release-notes {
-  color: var(--cp-color-text);
-  font-size: 12px;
-  font-weight: var(--font-weight-emphasis);
-  line-height: 1.65;
-  overflow-wrap: anywhere;
-}
-
-.release-notes :deep(*) {
-  max-width: 100%;
-}
-
-.release-notes :deep(:first-child) {
-  margin-top: 0;
-}
-
-.release-notes :deep(:last-child) {
-  margin-bottom: 0;
-}
-
-.release-notes :deep(h1),
-.release-notes :deep(h2),
-.release-notes :deep(h3),
-.release-notes :deep(h4) {
-  margin: 12px 0 6px;
-  color: var(--cp-color-text);
-  font-size: 12px;
-  font-weight: var(--font-weight-heavy);
-  line-height: 1.4;
-}
-
-.release-notes :deep(p) {
-  margin: 0 0 8px;
-}
-
-.release-notes :deep(ul),
-.release-notes :deep(ol) {
-  margin: 0 0 8px;
-  padding-left: 18px;
-}
-
-.release-notes :deep(li) {
-  margin: 3px 0;
-}
-
-.release-notes :deep(a) {
-  color: var(--cp-color-link);
-  font-weight: 700;
-  text-decoration: none;
-}
-
-.release-notes :deep(a:hover) {
-  color: var(--cp-color-link-hover);
-}
-
-.release-notes :deep(code) {
-  border-radius: 5px;
-  background: var(--cp-color-fill-tertiary);
-  color: var(--cp-color-text);
-  font-family: var(--font-mono);
-  font-size: 11px;
-  padding: 1px 5px;
-}
-
-.release-notes :deep(pre) {
-  margin: 8px 0;
-  overflow-x: auto;
-  border-radius: var(--cp-border-radius);
-  background: var(--cp-color-bg-container);
-  padding: 8px 10px;
-}
-
-.release-notes :deep(pre code) {
-  background: transparent;
-  padding: 0;
-}
-
-.release-notes :deep(blockquote) {
-  margin: 8px 0;
-  border-left: 3px solid var(--cp-color-split);
-  color: var(--cp-color-text-secondary);
-  padding-left: 10px;
-}
-</style>

frontend/src/router/routes.ts

@@ -55,6 +55,16 @@ export const routes: RouteRecordRaw[] = [
         name: 'settings',
         component: () => import('@/views/settings/index.vue'),
       },
+      {
+        path: 'settings/upstream',
+        name: 'settings-upstream',
+        component: () => import('@/views/settings/index.vue'),
+      },
+      {
+        path: 'settings/access',
+        name: 'settings-access',
+        component: () => import('@/views/settings/index.vue'),
+      },
       {
         path: 'settings/backup',
         name: 'settings-backup',

frontend/src/views/accounts/components/AccountCreateModal/AccountSetupFields.vue

@@ -43,7 +43,7 @@ const form = defineModel<AccountCreateForm>({ required: true })
         v-model="form.notes"
         :rows="3"
         :maxlength="500"
-        placeholder="最多 500 字，可不填。"
+        placeholder="最多 500 字，可不填"
         :disabled="disabled"
       />
     </BaseFormItem>

frontend/src/views/accounts/components/AccountCreateModal/presenter.ts

@@ -84,7 +84,7 @@ function resolveModal(
   else if (input.form.mode === 'oauth')
     description = '通过浏览器授权导入 OpenAI 账号'
   else if (input.form.mode === 'access_token')
-    description = '逐行粘贴 Access Token；未包含 Refresh Token 时无法自动续期'
+    description = '逐行粘贴 Access Token，未包含 Refresh Token 时无法自动续期'
   else if (input.form.mode === 'refresh_token')
     description = '逐行粘贴 Refresh Token，导入时将自动换取 Access Token'
 

frontend/src/views/accounts/components/AccountEditModal.vue

@@ -96,7 +96,7 @@ const selectedGroupIds = defineModel<string[]>('selectedGroupIds', { required: t
           v-model="notes"
           :rows="3"
           :maxlength="500"
-          placeholder="最多 500 字，留空可清除备注。"
+          placeholder="最多 500 字，留空可清除备注"
           :disabled="saving"
         />
       </BaseFormItem>

frontend/src/views/accounts/components/AccountProfileModal/TokenActivity.vue

@@ -47,7 +47,7 @@ const monthLabels = computed(() =>
     <BaseEmpty
       v-if="dailyUsage === null"
       title="暂无 Token 活动"
-      description="本次未获取到每日 Token 活动。"
+      description="本次未获取到每日 Token 活动"
       size="sm"
       surface="none"
     />

frontend/src/views/accounts/components/AccountProfileModal/index.vue

@@ -47,7 +47,7 @@ const { profile, subscription, loading, error, load } = useAccountPersonalInfo(a
         <BaseEmpty
           v-if="profile.hasStatsError"
           title="个人统计暂不可用"
-          description="本次未获取到统计数据，可以点击刷新信息重试。"
+          description="本次未获取到统计数据，可以点击刷新信息重试"
           :icon="TriangleAlert"
           surface="none"
         />
@@ -56,7 +56,7 @@ const { profile, subscription, loading, error, load } = useAccountPersonalInfo(a
           <AccountProfileActivityInsights :insights="profile.activityInsights" />
         </template>
         <p v-if="error" role="status" class="m-0 text-cp-sm text-cp-warning">
-          本次信息刷新未完成，保留已获取的结果。
+          本次信息刷新未完成，保留已获取的结果
         </p>
       </template>
     </div>

frontend/src/views/accounts/components/AccountQuotaForecastModal/index.vue

@@ -29,13 +29,13 @@ const { report, loading, refreshing, error, load, refresh } = useAccountQuotaFor
 )
 const { now, pause, resume } = useNow({ interval: 30_000, controls: true })
 const options = [
-  { label: '周额度', value: 'weekly' },
+  { label: '7日额度', value: 'weekly' },
   { label: '月额度', value: 'monthly' },
 ]
 const forecast = computed(() => report.value?.forecasts.find(item => item.period === period.value))
 const unavailableReason = computed(() => {
   if (forecast.value?.source && new Date(forecast.value.source.resetAt) <= now.value)
-    return '额度窗口已过期，请刷新账号额度后重试。'
+    return '额度窗口已过期，请刷新账号额度后重试'
   return forecast.value?.unavailableReason ?? null
 })
 
@@ -85,7 +85,7 @@ function handleExplanationKeydown(event: KeyboardEvent) {
       <div aria-live="polite" :aria-busy="loading || refreshing">
         <ForecastSkeleton v-if="loading || refreshing" />
         <div v-else-if="error && !report" class="grid rounded-cp-card bg-cp-fill-tertiary/70 [html[data-theme=light]_&]:bg-cp-fill-quaternary/70">
-          <BaseEmpty title="预测加载失败" description="暂时无法取得预测数据，请重新加载。" surface="none" class="min-h-80 content-center">
+          <BaseEmpty title="预测加载失败" description="暂时无法取得预测数据，请重新加载" surface="none" class="min-h-80 content-center">
             <template #action>
               <BaseButton variant="secondary" @click="load">
                 重新加载
@@ -122,16 +122,16 @@ function handleExplanationKeydown(event: KeyboardEvent) {
               仅供参考
             </h4>
             <p class="m-0">
-              按已记录数据估算，缺失的用量或费用可能使结果偏低。结果会随使用的模型和方式变化，并非官方承诺额度。
+              按已记录数据估算，缺失的用量或费用可能使结果偏低，结果会随使用的模型和方式变化，并非官方承诺额度
             </p>
             <p v-if="forecast?.lowSample" class="m-0">
-              目前数据还较少，结果可能有较大波动。
+              目前数据还较少，结果可能有较大波动
             </p>
             <p v-if="forecast?.extrapolated" class="m-0">
-              本页预测为折算值，剩余量仍按{{ forecast.source?.label ?? '当前周期' }}计算。
+              本页预测为折算值，剩余量仍按{{ forecast.source?.label ?? '当前周期' }}计算
             </p>
             <p class="m-0">
-              剩余量以数据更新时间为准；等价费用不是实际账单或账户余额。
+              剩余量以数据更新时间为准，等价费用不是实际账单或账户余额
             </p>
           </section>
         </BasePopover>

frontend/src/views/accounts/components/AccountQuotaPanel/ResetCredits.vue

@@ -156,7 +156,7 @@ function handleRequestConsume(creditId: string) {
             上次操作结果待确认
           </p>
           <p class="mt-1 mb-0 text-cp-xs leading-normal font-emphasis text-cp-text-secondary">
-            请继续确认上次重置结果。
+            请继续确认上次重置结果
           </p>
         </div>
         <BaseButton size="sm" variant="soft" :disabled="loading || consuming || !canRequestConsume" @click="requestConsume">
@@ -200,7 +200,7 @@ function handleRequestConsume(creditId: string) {
             class="mx-4 mt-0 mb-4 rounded-cp bg-cp-error-container px-4 py-3 text-cp-xs leading-normal font-emphasis text-cp-error-on-container"
             role="status"
           >
-            {{ loadError }}，请刷新重试。
+            {{ loadError }}，请刷新重试
           </p>
           <p
             v-else-if="loading && !hasSnapshot"

frontend/src/views/accounts/components/AccountQuotaSummaryCell/index.vue

@@ -40,7 +40,7 @@ const additionalEntryCount = computed(() => Math.max(summaryEntries.value.length
       </span>
       <div class="grid min-w-0 gap-1.5">
         <span class="text-[10px] leading-3 font-bold text-cp-text-quaternary">{{ account.usage.windowLabelDisplay }}</span>
-        <div class="h-1 w-full rounded-full bg-cp-success" title="上游额度未提供；绿色条不表示剩余额度" aria-hidden="true" />
+        <div class="h-1 w-full rounded-full bg-cp-success" title="上游额度未提供，绿色条不表示剩余额度" aria-hidden="true" />
       </div>
     </template>
     <template v-else-if="summaryEntries.length > 0">

frontend/src/views/accounts/components/AccountStatusBadge/presenter.ts

@@ -90,53 +90,53 @@ const displayDefinitions: Record<AccountStatusDisplayMode, AccountStatusDisplayD
     tone: 'warning',
     label: '退避中',
     title: 'OAuth 刷新退避',
-    description: '刷新失败，系统正在等待下一次自动尝试。',
-    recoveryHint: '系统会在计划时间自动重试。',
+    description: '刷新失败，系统正在等待下一次自动尝试',
+    recoveryHint: '系统会在计划时间自动重试',
     icon: Timer,
   },
   normal: {
     tone: statusTones.normal,
     label: statusLabels.normal,
-    description: '该账号当前状态需要关注。',
-    recoveryHint: '重新测试连接以获取最新状态。',
+    description: '该账号当前状态需要关注',
+    recoveryHint: '重新测试连接以获取最新状态',
     icon: CircleCheck,
   },
   quota_exhausted: {
     tone: statusTones.quota_exhausted,
     label: statusLabels.quota_exhausted,
-    description: '该账号当前状态需要关注。',
-    recoveryHint: '重新测试连接以获取最新状态。',
+    description: '该账号当前状态需要关注',
+    recoveryHint: '重新测试连接以获取最新状态',
     icon: Gauge,
   },
   rate_limited: {
     tone: statusTones.rate_limited,
     label: statusLabels.rate_limited,
-    description: '上游暂时限制了请求频率，系统正在冷却该账号。',
-    recoveryHint: '冷却结束后，系统会自动恢复调度。',
+    description: '上游暂时限制了请求频率，系统正在冷却该账号',
+    recoveryHint: '冷却结束后，系统会自动恢复调度',
     icon: Timer,
   },
   disabled: {
     tone: statusTones.disabled,
     label: statusLabels.disabled,
-    description: '该账号当前状态需要关注。',
-    recoveryHint: '重新测试连接以获取最新状态。',
+    description: '该账号当前状态需要关注',
+    recoveryHint: '重新测试连接以获取最新状态',
     icon: Power,
   },
   error: {
     tone: statusTones.error,
     label: statusLabels.error,
-    description: '该账号暂不参与调度，处理凭据后可重新测试连接。',
-    recoveryHint: '重新测试连接以获取最新状态。',
+    description: '该账号暂不参与调度，处理凭据后可重新测试连接',
+    recoveryHint: '重新测试连接以获取最新状态',
     icon: AlertTriangle,
   },
 }
 
 const errorRecoveryHints: Record<AccountErrorReason, string> = {
-  account_unverified: '重新授权后，账号会重新参与调度。',
-  access_token_expired: '重新授权后，账号会重新参与调度。',
-  credential_expired: '重新授权后，账号会重新参与调度。',
-  credential_invalid: '请更新或重新导入凭据，再次测试连接。',
-  account_banned: '请确认上游账号状态，解除限制后再启用。',
+  account_unverified: '重新授权后，账号会重新参与调度',
+  access_token_expired: '重新授权后，账号会重新参与调度',
+  credential_expired: '重新授权后，账号会重新参与调度',
+  credential_invalid: '请更新或重新导入凭据，再次测试连接',
+  account_banned: '请确认上游账号状态，解除限制后再启用',
 }
 
 export function resolveAccountStatusPresentation(
@@ -154,10 +154,10 @@ export function resolveAccountStatusPresentation(
   const definition = isFreeze
     ? {
         ...displayDefinitions.rate_limited,
-        description: '容量类请求失败累计达到阈值，系统已暂停该账号的调度。',
+        description: '容量类请求失败累计达到阈值，系统已暂停该账号的调度',
         recoveryHint: waitsForProbe
-          ? '冷却结束后进行恢复探测，成功后恢复调度；也可手动恢复账号。'
-          : '冷却结束后自动恢复调度，也可手动恢复账号。',
+          ? '冷却结束后进行恢复探测，成功后恢复调度，也可手动恢复账号'
+          : '冷却结束后自动恢复调度，也可手动恢复账号',
       }
     : displayDefinitions[mode]
   const nextRefreshDisplay = isBackoff ? formatDateTime(nextRefreshTimestamp) : null
@@ -175,8 +175,8 @@ export function resolveAccountStatusPresentation(
     ? errorRecoveryHints[input.errorReason]
     : definition.recoveryHint
   const hasDetail = input.status === 'error' || mode === 'rate_limited' || isBackoff
-  const retry = nextRefreshDisplay ? ` 下次尝试：${nextRefreshDisplay}。` : ''
-  const triggerLabel = `${title}。${definition.description}${retry} 点击或聚焦查看详情。`
+  const retry = nextRefreshDisplay ? `，下次尝试：${nextRefreshDisplay}` : ''
+  const triggerLabel = `${title}，${definition.description}${retry}，点击或聚焦查看详情`
 
   return {
     mode,

frontend/src/views/accounts/composables/useAccountResetCredits.ts

@@ -223,7 +223,7 @@ export function useAccountResetCredits(options: {
           ...operation,
           hasTransportFailure: true,
         }
-        toast.warning('消费结果暂不确定；重试会复用同一个请求标识', { duration: 5000 })
+        toast.warning('消费结果暂不确定，重试会复用同一个请求标识', { duration: 5000 })
       }
       else {
         target.pendingOperation = null

frontend/src/views/dashboard/components/HealthTimelinePointPopover.vue

@@ -126,7 +126,7 @@ const metricItems = outcomeMeta
       </div>
 
       <p class="m-0 border-t border-cp-split pt-3 text-[10px] leading-[1.45] font-emphasis text-cp-text-quaternary">
-        客户端取消与调用方错误单独记录，不计入可用性。
+        客户端取消与调用方错误单独记录，不计入可用性
       </p>
     </div>
   </section>

frontend/src/views/groups/components/AccountGroupFormModal.vue

@@ -9,7 +9,7 @@ import BaseFormItem from '@/components/base/BaseForm/FormItem.vue'
 import BaseForm from '@/components/base/BaseForm/index.vue'
 import BaseInput from '@/components/base/BaseInput.vue'
 import BaseModal from '@/components/base/BaseModal/index.vue'
-import BaseSwitch from '@/components/base/BaseSwitch.vue'
+import BaseSegmented from '@/components/base/BaseSegmented.vue'
 import BaseTextarea from '@/components/base/BaseTextarea.vue'
 import { ACCOUNT_GROUP_COLOR_PRESETS } from '../constants'
 
@@ -24,8 +24,8 @@ const open = defineModel<boolean>({ required: true })
 const form = defineModel<AccountGroupFormValue>('form', { required: true })
 const title = computed(() => props.group ? '编辑分组' : '创建分组')
 const description = computed(() => props.group
-  ? '修改分组名称、用途说明和 Fast 限制。'
-  : '创建后，可在账号管理中将账号加入这个分组。')
+  ? '修改分组名称、用途说明和 Fast 模式'
+  : '创建后，可在账号管理中将账号加入这个分组')
 </script>
 
 <template>
@@ -53,6 +53,19 @@ const description = computed(() => props.group
           :disabled="saving"
         />
       </BaseFormItem>
+      <BaseFormItem label="Fast 模式" description="关闭后按标准模式处理">
+        <BaseSegmented
+          :model-value="form.disableFast ? 'disabled' : 'default'"
+          class="w-48 max-w-full"
+          label="Fast 模式"
+          :options="[
+            { label: '默认', value: 'default' },
+            { label: '关闭', value: 'disabled' },
+          ]"
+          :disabled="saving"
+          @update:model-value="form.disableFast = $event === 'disabled'"
+        />
+      </BaseFormItem>
       <BaseFormItem label="描述（可选）">
         <BaseTextarea
           v-model="form.description"
@@ -62,14 +75,6 @@ const description = computed(() => props.group
           :disabled="saving"
         />
       </BaseFormItem>
-      <BaseFormItem description="开启后，绑定本分组的密钥将以普通模式处理 Fast 请求。任一绑定分组或全局关闭 Fast 时，限制均生效。">
-        <BaseSwitch
-          v-model="form.disableFast"
-          label="关闭 Fast"
-          show-label
-          :disabled="saving"
-        />
-      </BaseFormItem>
     </BaseForm>
 
     <template #footer>

frontend/src/views/groups/index.vue

@@ -187,14 +187,14 @@ const { allSelected, indeterminate, selectedRowKeys, toggleSelection, toggleAll
       @confirm="confirmBatchDelete"
     >
       <p class="m-0">
-        确定删除选中的 {{ selectedIds.size }} 个分组吗？账号本身不会被删除。
+        确定删除选中的 {{ selectedIds.size }} 个分组吗？账号本身不会被删除
       </p>
     </BaseConfirmModal>
 
     <BaseConfirmModal
       v-model="showDisableModal"
       title="禁用账号分组"
-      description="禁用后，使用该分组的 API 密钥将无法再使用其中的账号。"
+      description="禁用后，使用该分组的 API 密钥将无法再使用其中的账号"
       confirm-text="确认禁用"
       :loading="disabling"
       @confirm="confirmDisable"
@@ -206,7 +206,7 @@ const { allSelected, indeterminate, selectedRowKeys, toggleSelection, toggleAll
         将影响 {{ referencedKeyNames.length }} 个 API 密钥：{{ referencedKeyNames.join('、') }}
       </p>
       <p v-else-if="pendingDisableGroup?.clientKeyCount" class="mt-2 mb-0 text-cp-warning-text">
-        将影响 {{ pendingDisableGroup.clientKeyCount }} 个 API 密钥。
+        将影响 {{ pendingDisableGroup.clientKeyCount }} 个 API 密钥
       </p>
     </BaseConfirmModal>
 
@@ -220,7 +220,7 @@ const { allSelected, indeterminate, selectedRowKeys, toggleSelection, toggleAll
       @confirm="confirmDelete"
     >
       <p class="m-0">
-        确定删除“{{ pendingDeleteGroup?.name || '该分组' }}”吗？账号本身不会被删除。
+        确定删除“{{ pendingDeleteGroup?.name || '该分组' }}”吗？账号本身不会被删除
       </p>
     </BaseConfirmModal>
   </div>

frontend/src/views/key-usage/components/KeyUsageBudget.vue

@@ -8,7 +8,7 @@ import { keyUsageTime, money } from '../utils/format'
 const props = defineProps<{ budget: KeyUsageBudget }>()
 const windows = computed(() => [
   { label: '今日额度', limit: props.budget.dailyLimitUsd, used: props.budget.dailyUsedUsd, reset: props.budget.dailyResetsAt },
-  { label: '七日额度', limit: props.budget.weeklyLimitUsd, used: props.budget.weeklyUsedUsd, reset: props.budget.weeklyResetsAt },
+  { label: '7日额度', limit: props.budget.weeklyLimitUsd, used: props.budget.weeklyUsedUsd, reset: props.budget.weeklyResetsAt },
 ].map(window => ({
   ...window,
   limited: Number(window.limit) > 0,

frontend/src/views/key-usage/components/KeyUsageRecords.vue

@@ -39,7 +39,7 @@ const columns = computed(() => defineTableColumns<KeyUsageRecord>([
       <BaseSegmented v-model="kind" label="请求结果" :options="[{ label: '成功请求', value: 'success' }, { label: '错误记录', value: 'error' }]" />
     </template>
     <p v-if="error || stale" role="status" class="mt-0 mb-3 text-cp-sm text-cp-error-text">
-      {{ error || '请求日志刷新失败，暂时保留上次结果。' }}
+      {{ error || '请求日志刷新失败，暂时保留上次结果' }}
     </p>
     <div class="flex h-120 min-h-0 overflow-hidden">
       <BaseTable class="min-w-0 flex-1" :columns="columns" :rows="rows" :loading="loading" scrollbar-always-visible :empty-text="error ? '请求日志加载失败，请点击顶部刷新重试' : '所选条件下暂无记录'">

frontend/src/views/key-usage/components/KeyUsageSkeleton.vue

@@ -56,7 +56,7 @@ const metrics = [
     <BaseCard title="额度概览" aria-busy="true">
       <div class="flex flex-1 flex-col justify-between gap-6" aria-hidden="true">
         <div class="grid flex-1 gap-6 sm:grid-cols-2">
-          <div v-for="label in ['今日额度', '七日额度']" :key="label" class="flex min-w-0 flex-col justify-between gap-4">
+          <div v-for="label in ['今日额度', '7日额度']" :key="label" class="flex min-w-0 flex-col justify-between gap-4">
             <div>
               <div class="text-cp-sm text-cp-text-secondary">
                 {{ label }}

frontend/src/views/key-usage/index.vue

@@ -30,7 +30,7 @@ const { items, currentPage, pageSize, total, loading: recordsLoading, error: rec
           </BaseInput>
         </div>
         <p v-if="overviewError" role="alert" class="m-0 rounded-cp-lg bg-cp-error-container px-4 py-3 text-cp-sm text-cp-error-text">
-          {{ overviewError }}{{ overview ? '，暂时保留上次结果。' : '，请点击顶部刷新重试。' }}
+          {{ overviewError }}{{ overview ? '，暂时保留上次结果' : '，请点击顶部刷新重试' }}
         </p>
         <template v-if="overview">
           <KeyUsageSummary :summary="overview.summary" />

frontend/src/views/keys/components/ApiKeyBudgetCell.vue

@@ -7,7 +7,7 @@ import { formatDateTime } from '@/utils/date'
 const props = defineProps<{ apiKey: ApiKey }>()
 const windows = computed(() => [
   { label: '日', heading: '日用量', used: props.apiKey.dailyUsedUsd, limit: props.apiKey.dailyLimitUsd, reset: props.apiKey.dailyResetsAt },
-  { label: '周', heading: '周用量', used: props.apiKey.weeklyUsedUsd, limit: props.apiKey.weeklyLimitUsd, reset: props.apiKey.weeklyResetsAt },
+  { label: '7日', heading: '7日用量', used: props.apiKey.weeklyUsedUsd, limit: props.apiKey.weeklyLimitUsd, reset: props.apiKey.weeklyResetsAt },
 ])
 function amount(value: string) {
   // 列表最多显示两位小数，不补末尾零；明细保留原始金额的全部精度。
@@ -26,7 +26,7 @@ function amount(value: string) {
         aria-haspopup="dialog"
       >
         <span v-for="window in windows" :key="window.label" class="flex min-w-0 items-center gap-2">
-          <span class="shrink-0 text-cp-text-tertiary">{{ window.label }}</span>
+          <span class="w-6 shrink-0 text-cp-text-tertiary">{{ window.label }}</span>
           <span class="truncate" :class="Number(window.limit) > 0 && Number(window.used) >= Number(window.limit) ? 'text-cp-error' : 'text-cp-text'">
             ${{ amount(window.used) }} / {{ Number(window.limit) === 0 ? '∞' : `$${amount(window.limit)}` }}
           </span>

frontend/src/views/keys/components/ApiKeyBudgetResetModal.vue

@@ -1,8 +1,10 @@
 <script setup lang="ts">
 import type { ApiKey } from '@/api'
-import { shallowRef, watch } from 'vue'
+import { ArrowRight } from '@lucide/vue'
+import { computed, shallowRef, watch } from 'vue'
 import { resetApiKeyBudget } from '@/api'
-import BaseConfirmModal from '@/components/base/BaseConfirmModal.vue'
+import BaseButton from '@/components/base/BaseButton.vue'
+import BaseModal from '@/components/base/BaseModal/index.vue'
 import BaseSegmented from '@/components/base/BaseSegmented.vue'
 import { toast } from '@/components/base/BaseToast'
 import { useAsyncAction } from '@/composables/useAsyncAction'
@@ -15,8 +17,14 @@ const { loading, run } = useAsyncAction()
 const periods = [
   { label: '全部', value: 'all' },
   { label: '日额度', value: 'daily' },
-  { label: '周额度', value: 'weekly' },
+  { label: '7日额度', value: 'weekly' },
 ]
+const windows = computed(() => props.apiKey
+  ? [
+      { value: 'daily', label: '日额度', used: props.apiKey.dailyUsedUsd },
+      { value: 'weekly', label: '7日额度', used: props.apiKey.weeklyUsedUsd },
+    ].map(window => ({ ...window, selected: period.value === 'all' || period.value === window.value }))
+  : [])
 
 watch(open, (value) => {
   if (value)
@@ -38,23 +46,56 @@ async function resetBudget() {
 </script>
 
 <template>
-  <BaseConfirmModal
+  <BaseModal
     v-model="open"
     title="重置已用额度"
-    description="选择要清零的日额度、周额度或全部额度"
-    confirm-text="确认重置"
-    :loading="loading"
-    :confirm-disabled="!apiKey"
-    @confirm="resetBudget"
+    description="清零所选周期的已用金额，原重置时间保持不变"
+    size="sm"
+    :dismissible="!loading"
   >
-    <div class="space-y-4">
-      <p class="m-0 break-words text-cp-text">
-        {{ apiKey?.name }}
-      </p>
-      <BaseSegmented v-model="period" class="w-full" label="重置范围" :options="periods" :disabled="loading" />
-      <p class="m-0 text-cp leading-relaxed text-cp-text-secondary">
-        限额上限、原重置时间和历史使用记录保持不变，正在进行的请求会在完成后计入额度
-      </p>
+    <div class="grid gap-5">
+      <div class="grid min-w-0 gap-1.5">
+        <span class="text-cp-xs text-cp-text-quaternary">目标密钥</span>
+        <p class="m-0 break-words text-cp text-cp-text">
+          {{ apiKey?.name }}
+        </p>
+      </div>
+
+      <fieldset class="m-0 min-w-0 border-0 p-0" :disabled="loading">
+        <legend class="mb-2.5 p-0 text-cp-sm text-cp-text-secondary">
+          重置范围
+        </legend>
+        <BaseSegmented v-model="period" class="w-full" label="重置范围" :options="periods" :disabled="loading" />
+      </fieldset>
+
+      <div v-if="apiKey" class="rounded-cp bg-cp-fill-quaternary p-3.5" aria-live="polite" aria-atomic="true">
+        <div class="mb-3 grid grid-cols-[3.5rem_minmax(0,1fr)_1rem_4.5rem] items-center gap-2 text-cp-xs text-cp-text-quaternary">
+          <span class="col-span-2 text-right">当前已用</span>
+          <span class="col-start-4 text-right">重置后</span>
+        </div>
+        <dl class="m-0 grid gap-3">
+          <div v-for="window in windows" :key="window.value" class="grid grid-cols-[3.5rem_minmax(0,1fr)_1rem_4.5rem] items-center gap-2">
+            <dt class="text-cp-sm text-cp-text-secondary">
+              {{ window.label }}
+            </dt>
+            <dd class="col-span-3 m-0 grid grid-cols-subgrid items-center">
+              <span class="min-w-0 break-all text-right font-mono text-cp-sm text-cp-text tabular-nums">${{ window.used }}</span>
+              <ArrowRight class="size-3.5 text-cp-text-quaternary" aria-hidden="true" />
+              <span v-if="window.selected" class="text-right font-mono text-cp-sm text-cp-primary-text tabular-nums">$0</span>
+              <span v-else class="text-right text-cp-xs text-cp-text-quaternary">保持不变</span>
+            </dd>
+          </div>
+        </dl>
+      </div>
     </div>
-  </BaseConfirmModal>
+
+    <template #footer>
+      <BaseButton variant="secondary" :disabled="loading" @click="open = false">
+        取消
+      </BaseButton>
+      <BaseButton variant="primary" :loading="loading" :disabled="!apiKey" @click="resetBudget">
+        确认重置
+      </BaseButton>
+    </template>
+  </BaseModal>
 </template>

frontend/src/views/keys/components/ApiKeyCreateModal.vue

@@ -28,13 +28,14 @@ const emit = defineEmits<{
 const open = defineModel<boolean>({ default: false })
 const createdOpen = defineModel<boolean>('createdOpen', { default: false })
 const form = defineModel<ApiKeyFormValue>('form', { required: true })
-const title = computed(() => props.editing ? '编辑 API Key' : '创建 API Key')
+const title = computed(() => props.editing ? '编辑密钥' : '创建 API Key')
 </script>
 
 <template>
   <BaseModal
     v-model="open"
     :title="title"
+    description="配置密钥信息、分组与使用限制"
     tone="info"
     size="lg"
     :dismissible="!saving"
@@ -106,13 +107,13 @@ const title = computed(() => props.editing ? '编辑 API Key' : '创建 API Key'
             </template>
           </BaseInput>
         </BaseFormItem>
-        <BaseFormItem label="周限额">
+        <BaseFormItem label="7日限额">
           <BaseInput
             v-model="form.weeklyLimitUsd"
             type="number"
             min="0"
             step="any"
-            aria-label="周限额（美元）"
+            aria-label="7日限额（美元）"
             placeholder="不限制"
             :disabled="saving"
           >

frontend/src/views/keys/components/ApiKeyIdentityCell.vue

@@ -9,12 +9,12 @@ defineProps<{
 </script>
 
 <template>
-  <div class="flex min-w-0 flex-col gap-0.5">
-    <span class="truncate text-cp font-bold text-cp-text" :title="apiKey.name">
+  <div class="grid min-w-0 gap-1">
+    <span class="truncate text-cp text-cp-text" :title="apiKey.name">
       {{ apiKey.name }}
     </span>
-    <span v-if="apiKey.label" class="truncate text-cp-sm font-emphasis text-cp-text-tertiary" :title="apiKey.label">
-      {{ apiKey.label }}
+    <span class="truncate text-cp-xs font-emphasis text-cp-text-quaternary" :title="apiKey.label || undefined">
+      {{ apiKey.label || '未填写备注' }}
     </span>
   </div>
 </template>

frontend/src/views/keys/components/ApiKeyUseModal.vue

@@ -77,7 +77,7 @@ const visibleFiles = computed(() => [
       </div>
 
       <p class="text-cp-sm text-cp-text-secondary">
-        配置含密钥，请勿分享，覆盖前请备份。
+        配置含密钥，请勿分享，覆盖前请备份
       </p>
 
       <div class="flex flex-col gap-3">

frontend/src/views/keys/composables/useApiKeyMutations.ts

@@ -138,7 +138,7 @@ export function useApiKeyMutations(options: {
   }
 
   function validateForm() {
-    for (const [label, value] of [['日限额', form.value.dailyLimitUsd], ['周限额', form.value.weeklyLimitUsd]]) {
+    for (const [label, value] of [['日限额', form.value.dailyLimitUsd], ['7日限额', form.value.weeklyLimitUsd]]) {
       if (value.trim() && !/^\d{1,10}(?:\.\d{1,10})?$/.test(value.trim())) {
         toast.warning(`${label}必须是非负金额，最多 10 位小数`)
         return false

frontend/src/views/keys/index.vue

@@ -246,13 +246,13 @@ watch(
     <BaseConfirmModal
       v-model="showAllAccountsConfirm"
       title="授予全部账号权限"
-      description="保存后，该密钥可以使用所有账号。"
+      description="保存后，该密钥可以使用所有账号"
       confirm-text="确认授予全部账号"
       :loading="savingKey"
       @confirm="confirmAllAccountsScope"
     >
       <p class="m-0">
-        该密钥可以使用所有账号，包括以后新增和未分组的账号。
+        该密钥可以使用所有账号，包括以后新增和未分组的账号
       </p>
     </BaseConfirmModal>
 

frontend/src/views/proxies/components/ProxyAccountsModal.vue

@@ -125,7 +125,7 @@ watch([open, () => props.proxy?.id], () => {
   </BaseModal>
   <BaseConfirmModal v-model="showRemove" title="移除关联账号" confirm-text="移除" :loading="removing" @confirm="confirmRemove">
     <p class="m-0 break-words">
-      将“{{ pendingRemove?.name }}”从“{{ proxy?.name }}”移除后，该账号将改为直连。
+      将“{{ pendingRemove?.name }}”从“{{ proxy?.name }}”移除后，该账号将改为直连
     </p>
   </BaseConfirmModal>
 </template>

frontend/src/views/proxies/components/ProxyFormModal.vue

@@ -29,8 +29,8 @@ const showSecret = shallowRef(false)
 const busy = computed(() => props.saving || props.testing)
 const title = computed(() => props.proxy ? '编辑代理' : '新增代理')
 const connectionDescription = computed(() => props.proxy
-  ? '留空保留当前连接和认证信息；填写新地址时，请包含所需的用户名和密码。'
-  : '支持 HTTP、HTTPS、SOCKS5 和 SOCKS5H，可在地址中包含用户名和密码。')
+  ? '留空保留当前连接和认证信息，填写新地址时，请包含所需的用户名和密码'
+  : '支持 HTTP、HTTPS、SOCKS5 和 SOCKS5H，可在地址中包含用户名和密码')
 
 watch(open, () => {
   showSecret.value = false
@@ -63,7 +63,7 @@ watch(open, () => {
       <BaseSwitch v-model="customLocation" label="自定义时区位置" show-label :disabled="busy" />
       <RequestLocationFields v-if="customLocation" v-model="location" :disabled="busy" />
       <p v-if="proxy?.accountCount && proxyUrl.trim()" class="m-0 text-cp-sm text-cp-warning-text">
-        将更新 {{ proxy.accountCount }} 个关联账号的出口。
+        将更新 {{ proxy.accountCount }} 个关联账号的出口
       </p>
     </BaseForm>
     <template #footer>

frontend/src/views/proxies/index.vue

@@ -212,7 +212,7 @@ onMounted(() => void query.execute())
         <div class="flex h-full min-h-0 flex-col">
           <BaseTable class="min-h-0 flex-1" :columns="columns" :rows="proxies" :loading="loading" :empty-text="search.trim() ? '没有找到匹配的代理，请尝试其他名称' : '暂无代理，请点击新增代理添加'">
             <template #name="{ row }">
-              <strong class="block truncate text-cp text-cp-text" :title="row.name">{{ row.name }}</strong>
+              <span class="block truncate text-cp text-cp-text" :title="row.name">{{ row.name }}</span>
             </template>
             <template #address="{ row }">
               <div class="grid min-w-0 gap-1">

frontend/src/views/settings/components/AccountAutoFreezeCard.vue

@@ -22,25 +22,25 @@ const adaptiveConcurrencyHintId = useId()
 
 <template>
   <BaseCard
-    title="过载策略"
-    description="上游繁忙或服务不可用时，使用过载策略保护"
+    title="过载保护"
+    description="上游容量类错误累计达到阈值时，暂停账号调度并进入冷却"
   >
     <BaseForm class="max-w-6xl sm:grid-cols-2">
       <BaseSwitch
         v-model="enabled"
         class="col-span-full justify-self-start"
-        label="启用策略"
+        label="启用过载保护"
         show-label
       />
 
       <BaseFormItem
-        label="触发阈值"
-        description="窗口内累计失败尝试达到此次数后触发"
+        label="失败次数阈值"
+        description="统计窗口内累计容量类失败达到此次数后触发"
       >
         <BaseInput
           v-model="threshold"
           :disabled="!enabled"
-          aria-label="触发阈值"
+          aria-label="失败次数阈值"
           type="number"
           min="2"
           max="1000"

frontend/src/views/settings/components/AdminPasswordCard.vue

@@ -3,6 +3,7 @@ import { LockKeyhole } from '@lucide/vue'
 import { shallowRef } from 'vue'
 import BaseButton from '@/components/base/BaseButton.vue'
 import BaseCard from '@/components/base/BaseCard.vue'
+import BaseMotionIcon from '@/components/base/BaseMotionIcon.vue'
 import ChangePasswordModal from './ChangePasswordModal.vue'
 
 const open = shallowRef(false)
@@ -13,14 +14,25 @@ const open = shallowRef(false)
     <template #actions>
       <BaseButton variant="secondary" @click="open = true">
         <template #icon>
-          <LockKeyhole class="size-4" />
+          <LockKeyhole class="size-4" aria-hidden="true" />
         </template>
         修改密码
       </BaseButton>
     </template>
-    <p class="m-0 text-cp leading-relaxed text-cp-text-secondary">
-      修改后，所有已登录的管理员都需要使用新密码重新登录。
-    </p>
+
+    <div class="flex min-h-16 max-w-6xl items-center gap-3 rounded-cp bg-cp-fill-quaternary px-4 py-3">
+      <BaseMotionIcon class="inline-flex size-9 shrink-0 items-center justify-center rounded-cp bg-cp-bg-container text-cp-primary-text shadow-cp-tertiary">
+        <LockKeyhole class="size-4" aria-hidden="true" />
+      </BaseMotionIcon>
+      <div class="min-w-0">
+        <p class="m-0 text-cp leading-[1.15] font-bold text-cp-text">
+          验证当前密码
+        </p>
+        <p class="mt-1.5 mb-0 text-cp-sm leading-snug font-emphasis text-cp-text-secondary">
+          修改后，所有管理员需重新登录
+        </p>
+      </div>
+    </div>
   </BaseCard>
   <ChangePasswordModal v-model="open" />
 </template>

frontend/src/views/settings/components/ChangePasswordModal.vue

@@ -63,9 +63,6 @@ async function submit() {
       <FormItem label="确认新密码" :error="confirmationError" required>
         <BaseInput v-model="form.confirmation" type="password" autocomplete="new-password" placeholder="再次输入新密码" :disabled="loading" maxlength="1024" />
       </FormItem>
-      <p class="m-0 rounded-cp bg-cp-fill-quaternary px-3 py-2.5 text-cp-sm leading-relaxed text-cp-text-secondary">
-        保存后，当前及其他设备上的管理员登录都会失效，需要重新登录。
-      </p>
     </form>
     <template #footer>
       <BaseButton variant="secondary" :disabled="loading" @click="open = false">

frontend/src/views/settings/components/FastPolicyCard.vue

@@ -1,20 +0,0 @@
-<script setup lang="ts">
-import BaseCard from '@/components/base/BaseCard.vue'
-import BaseFormItem from '@/components/base/BaseForm/FormItem.vue'
-import BaseForm from '@/components/base/BaseForm/index.vue'
-import BaseSwitch from '@/components/base/BaseSwitch.vue'
-
-defineProps<{ disabled: boolean }>()
-
-const model = defineModel<boolean>({ required: true })
-</script>
-
-<template>
-  <BaseCard title="服务档位" description="Fast 模式下的请求将按普通档处理，关闭后仍遵循分组限制">
-    <BaseForm class="max-w-6xl">
-      <BaseFormItem>
-        <BaseSwitch v-model="model" label="关闭" show-label :disabled="disabled" />
-      </BaseFormItem>
-    </BaseForm>
-  </BaseCard>
-</template>

frontend/src/views/settings/components/ModelAliasesCard.vue

@@ -25,7 +25,7 @@ const emit = defineEmits<{
 <template>
   <BaseCard
     title="模型映射"
-    description="配置请求模型与上游模型的映射关系"
+    description="将客户端请求的模型名称映射为实际使用的上游模型"
   >
     <div class="grid gap-4">
       <div class="flex flex-wrap items-center gap-3">

frontend/src/views/settings/components/RequestLocationCard.vue

@@ -11,10 +11,10 @@ const enabled = defineModel<boolean>('enabled', { required: true })
 </script>
 
 <template>
-  <BaseCard title="请求时区与位置" description="开启后覆盖 Codex 请求的时区与位置，代理自定义配置优先">
+  <BaseCard title="请求位置覆盖" description="覆盖 Codex 请求中的时区与位置字段，账号出站代理的自定义位置优先">
     <template #body>
       <div>
-        <BaseSwitch v-model="enabled" label="启用" show-label :disabled="disabled" />
+        <BaseSwitch v-model="enabled" label="启用位置覆盖" show-label :disabled="disabled" />
       </div>
       <BaseForm class="mt-4 max-w-6xl">
         <RequestLocationFields v-model="location" :disabled="disabled || !enabled" />

frontend/src/views/settings/components/RequestQueueCard.vue

@@ -10,16 +10,16 @@ const concurrencyWaitTimeoutSeconds = defineModel<string>('concurrencyWaitTimeou
 </script>
 
 <template>
-  <BaseCard title="请求排队">
+  <BaseCard title="并发队列">
     <BaseForm class="max-w-6xl sm:grid-cols-2">
-      <BaseFormItem label="每个密钥最大排队数" description="各密钥独立计数；0 表示不排队">
-        <BaseInput v-model="maxWaitingPerKey" aria-label="每个密钥最大排队数" type="number" min="0" max="1000" step="1" />
+      <BaseFormItem label="密钥队列容量" description="每个客户端密钥允许等待的最大请求数，0 表示不排队">
+        <BaseInput v-model="maxWaitingPerKey" aria-label="密钥队列容量" type="number" min="0" max="1000" step="1" />
       </BaseFormItem>
-      <BaseFormItem label="单账号最大排队数" description="各账号独立计数；0 表示不排队">
-        <BaseInput v-model="maxWaitingPerAccount" aria-label="单账号最大排队数" type="number" min="0" max="1000" step="1" />
+      <BaseFormItem label="账号队列容量" description="每个上游账号允许等待的最大请求数，0 表示不排队">
+        <BaseInput v-model="maxWaitingPerAccount" aria-label="账号队列容量" type="number" min="0" max="1000" step="1" />
       </BaseFormItem>
-      <BaseFormItem label="最长排队秒数" description="密钥和账号共用，从首次入队起计时，1～120 秒">
-        <BaseInput v-model="concurrencyWaitTimeoutSeconds" aria-label="最长排队秒数" type="number" min="1" max="120" step="1" />
+      <BaseFormItem label="排队超时（秒）" description="密钥队列与账号队列共用的等待时限，从首次入队起计时，1～120 秒">
+        <BaseInput v-model="concurrencyWaitTimeoutSeconds" aria-label="排队超时（秒）" type="number" min="1" max="120" step="1" />
       </BaseFormItem>
     </BaseForm>
   </BaseCard>

frontend/src/views/settings/components/ResponseBodyLimitCard.vue

@@ -8,10 +8,10 @@ const model = defineModel<string>({ required: true })
 </script>
 
 <template>
-  <BaseCard title="请求体解压上限">
+  <BaseCard title="请求体限制" description="限制 Responses 请求解压后的请求体大小">
     <BaseForm class="max-w-6xl sm:grid-cols-2 sm:items-end">
-      <BaseFormItem label="最大解压大小" description="调高可接收更大的图片和历史上下文，也会增加内存占用">
-        <BaseInput v-model="model" aria-label="请求体解压上限" type="number" min="1" step="1">
+      <BaseFormItem label="解压后大小上限" description="调高可接收更大的图片和历史上下文，也会增加内存占用">
+        <BaseInput v-model="model" aria-label="解压后请求体大小上限" type="number" min="1" step="1">
           <template #suffix>
             MiB
           </template>

frontend/src/views/settings/components/RuntimeSettingsCard.vue

@@ -11,15 +11,15 @@ const requestIntervalMs = defineModel<string>('requestIntervalMs', { required: t
 </script>
 
 <template>
-  <BaseCard title="并发与请求间隔">
+  <BaseCard title="并发控制">
     <BaseForm class="max-w-6xl sm:grid-cols-2">
       <BaseFormItem
-        label="单账号默认最大并发"
+        label="默认账号并发上限"
         description="账号未单独设置时使用的并发上限"
       >
         <BaseInput
           v-model="maxConcurrentPerAccount"
-          aria-label="单账号默认最大并发"
+          aria-label="默认账号并发上限"
           type="number"
         >
           <template #prefix>
@@ -29,19 +29,19 @@ const requestIntervalMs = defineModel<string>('requestIntervalMs', { required: t
       </BaseFormItem>
 
       <BaseFormItem
-        label="请求间隔"
+        label="最小请求间隔"
         description="控制同一账号两次调度之间的最小等待时间"
       >
         <BaseInput
           v-model="requestIntervalMs"
-          aria-label="请求间隔 ms"
+          aria-label="最小请求间隔（毫秒）"
           type="number"
         >
           <template #prefix>
             <Timer class="size-4" />
           </template>
           <template #suffix>
-            ms
+            毫秒
           </template>
         </BaseInput>
       </BaseFormItem>

frontend/src/views/settings/components/SettingsAccessSection.vue

@@ -0,0 +1,77 @@
+<script setup lang="ts">
+import { onMounted } from 'vue'
+import BaseConfirmModal from '@/components/base/BaseConfirmModal.vue'
+import { useAdminApiKey } from '../composables/useAdminApiKey'
+import AdminApiKeyCard from './AdminApiKeyCard.vue'
+import AdminPasswordCard from './AdminPasswordCard.vue'
+import ClientVersionSettings from './client-version/index.vue'
+import ResponseBodyLimitCard from './ResponseBodyLimitCard.vue'
+
+defineProps<{
+  disabled: boolean
+  loading: boolean
+  desktopError: string
+  cliError: string
+}>()
+
+const minCodexDesktopVersion = defineModel<string>('minCodexDesktopVersion', { required: true })
+const minCodexCliVersion = defineModel<string>('minCodexCliVersion', { required: true })
+const responsesMaxDecompressedBodyMiB = defineModel<string>('responsesMaxDecompressedBodyMiB', { required: true })
+
+const {
+  loading: adminKeyLoading,
+  regenerating,
+  deleting,
+  showDeleteModal,
+  generatedKey,
+  status,
+  regenerate,
+  remove,
+  copyGeneratedKey,
+  loadStatus,
+} = useAdminApiKey()
+
+onMounted(loadStatus)
+</script>
+
+<template>
+  <div class="grid min-w-0 gap-5">
+    <AdminPasswordCard />
+
+    <AdminApiKeyCard
+      :status="status"
+      :loading="adminKeyLoading"
+      :regenerating="regenerating"
+      :deleting="deleting"
+      :generated-key="generatedKey"
+      @regenerate="regenerate"
+      @request-delete="showDeleteModal = true"
+      @copy="copyGeneratedKey"
+    />
+
+    <fieldset :disabled="disabled" class="m-0 grid min-w-0 gap-5 border-0 p-0" aria-label="访问限制">
+      <ClientVersionSettings
+        v-model:min-codex-desktop-version="minCodexDesktopVersion"
+        v-model:min-codex-cli-version="minCodexCliVersion"
+        :loading="loading"
+        :desktop-error="desktopError"
+        :cli-error="cliError"
+      />
+      <ResponseBodyLimitCard v-model="responsesMaxDecompressedBodyMiB" />
+    </fieldset>
+
+    <BaseConfirmModal
+      v-model="showDeleteModal"
+      title="删除管理员 API Key"
+      description="删除后外部系统将无法继续使用该 Key 调用管理接口"
+      destructive
+      confirm-text="确认删除"
+      :loading="deleting"
+      @confirm="remove"
+    >
+      <p class="m-0">
+        确定要删除当前管理员 API Key 吗？此操作会立即生效
+      </p>
+    </BaseConfirmModal>
+  </div>
+</template>

frontend/src/views/settings/components/TokenRefreshCard.vue

@@ -10,17 +10,17 @@ const refreshConcurrency = defineModel<string>('refreshConcurrency', { required:
 </script>
 
 <template>
-  <BaseCard title="凭据自动刷新">
+  <BaseCard title="凭据刷新">
     <BaseForm class="max-w-6xl sm:grid-cols-2">
-      <BaseFormItem label="提前刷新秒数" description="Token 过期前多少秒触发刷新">
-        <BaseInput v-model="refreshMarginSeconds" aria-label="提前刷新秒数" type="number">
+      <BaseFormItem label="提前刷新时间（秒）" description="访问令牌到期前提前触发自动刷新的时间">
+        <BaseInput v-model="refreshMarginSeconds" aria-label="提前刷新时间（秒）" type="number">
           <template #prefix>
             <Timer class="size-4" />
           </template>
         </BaseInput>
       </BaseFormItem>
-      <BaseFormItem label="刷新并发数" description="同时刷新 Token 的最大请求数，减小可避免限流">
-        <BaseInput v-model="refreshConcurrency" aria-label="刷新并发数" type="number">
+      <BaseFormItem label="刷新并发上限" description="同时执行凭据刷新的最大请求数，降低上限可减少限流风险">
+        <BaseInput v-model="refreshConcurrency" aria-label="刷新并发上限" type="number">
           <template #prefix>
             <Zap class="size-4" />
           </template>

frontend/src/views/settings/components/backup/BackupScheduleCard.vue

@@ -40,8 +40,8 @@ const TIMEZONE_OPTIONS = [
 
 <template>
   <BaseCard
-    title="定时备份"
-    description="配置自动定时备份"
+    title="备份计划"
+    description="配置自动备份的执行时间与保留策略"
   >
     <template #actions>
       <BaseButton variant="primary" :loading="saving" :disabled="loading" @click="emit('save')">
@@ -63,7 +63,7 @@ const TIMEZONE_OPTIONS = [
           />
         </div>
 
-        <BaseFormItem label="时区" description="显式 IANA 时区">
+        <BaseFormItem label="调度时区" description="用于解释 Cron 表达式的 IANA 时区">
           <BaseSelect v-model="schedule.scheduleTimezone" :options="TIMEZONE_OPTIONS" />
         </BaseFormItem>
 
@@ -78,13 +78,13 @@ const TIMEZONE_OPTIONS = [
           </BaseInput>
         </BaseFormItem>
 
-        <BaseFormItem label="备份过期天数" description="超过此天数自动删除，0 = 永不过期">
-          <BaseInput v-model="schedule.retentionDays" aria-label="备份过期天数" type="number" min="0" />
+        <BaseFormItem label="保留天数" description="超过此天数自动删除，0 表示不按天数清理，仍受最大保留份数限制">
+          <BaseInput v-model="schedule.retentionDays" aria-label="备份保留天数" type="number" min="0" />
         </BaseFormItem>
 
         <BaseFormItem
           label="最大保留份数"
-          description="最多保留的备份数量，0 = 不限制"
+          description="最多保留的备份数量，0 表示不按份数清理，仍受保留天数限制"
         >
           <BaseInput v-model="schedule.retentionCount" aria-label="最大保留份数" type="number" min="0" />
         </BaseFormItem>

frontend/src/views/settings/components/backup/BackupStorageCard.vue

@@ -104,7 +104,7 @@ function toggleSecretVisible(): void {
           </BaseInput>
         </BaseFormItem>
 
-        <BaseFormItem label="区域" description="R2 使用固定值 auto；其它服务按提供方填写">
+        <BaseFormItem label="区域" description="R2 使用固定值 auto，其它服务按提供方填写">
           <BaseInput v-model="storage.region" aria-label="区域" />
         </BaseFormItem>
 
@@ -112,8 +112,8 @@ function toggleSecretVisible(): void {
           <BaseInput v-model="storage.bucket" aria-label="存储桶" />
         </BaseFormItem>
 
-        <BaseFormItem label="Key 前缀" description="对象前缀，历史归档已保存完整旧前缀">
-          <BaseInput v-model="storage.prefix" aria-label="Key 前缀" />
+        <BaseFormItem label="对象键前缀" description="备份对象的存储路径前缀，不影响已有备份">
+          <BaseInput v-model="storage.prefix" aria-label="对象键前缀" />
         </BaseFormItem>
 
         <BaseFormItem label="Access Key ID" description="对象存储专用凭据">
@@ -161,7 +161,7 @@ function toggleSecretVisible(): void {
         </BaseFormItem>
 
         <div class="col-span-2 flex items-center gap-4 @max-[640px]:col-span-1">
-          <BaseCheckbox v-model="storage.forcePathStyle" label="强制路径风格" show-label />
+          <BaseCheckbox v-model="storage.forcePathStyle" label="强制路径式访问" show-label />
         </div>
       </BaseForm>
     </div>

frontend/src/views/settings/components/backup/R2GuideModal.vue

@@ -18,8 +18,8 @@ const open = defineModel<boolean>({ default: false })
         </h3>
         <p class="m-0">
           在 Cloudflare Dashboard 进入 R2，创建一个私有存储桶（例如
-          <code class="rounded bg-cp-fill-quaternary px-1.5 py-0.5">codex-proxy-backups</code>）。
-          桶必须保持私有，不要开放公共读取。
+          <code class="rounded bg-cp-fill-quaternary px-1.5 py-0.5">codex-proxy-backups</code>），
+          桶必须保持私有，不要开放公共读取
         </p>
       </section>
 
@@ -33,7 +33,7 @@ const open = defineModel<boolean>({ default: false })
           <code class="rounded bg-cp-fill-quaternary px-1.5 py-0.5">对象写</code>、
           <code class="rounded bg-cp-fill-quaternary px-1.5 py-0.5">对象查看</code> 与
           <code class="rounded bg-cp-fill-quaternary px-1.5 py-0.5">对象删除</code> 权限，
-          不要授予整个账号权限。
+          不要授予整个账号权限
         </p>
       </section>
 
@@ -56,8 +56,8 @@ const open = defineModel<boolean>({ default: false })
           4. 完成连接测试
         </h3>
         <p class="m-0">
-          保存配置后点击「测试连接」，系统会写入一个探针对象并校验读写权限。测试通过后
-          才能启用自动计划或手动创建备份。凭据只保存在服务端，不会写入浏览器。
+          保存配置后点击「测试连接」，系统会写入一个探针对象并校验读写权限，测试通过后
+          才能启用自动计划或手动创建备份，凭据只保存在服务端，不会写入浏览器
         </p>
       </section>
     </div>

frontend/src/views/settings/components/client-version/components/ClientCliInstallPanel.vue

@@ -40,7 +40,7 @@ const cliCommands = [
           安装或升级 Codex CLI
         </h3>
         <p class="mt-1.5 mb-0 text-cp-sm leading-[1.45] font-semibold text-cp-text-secondary">
-          按当前安装来源执行一种命令；官方安装器会自动选择当前平台架构。
+          按当前安装来源执行一种命令，官方安装器会自动选择当前平台架构
         </p>
       </div>
     </div>
@@ -53,7 +53,7 @@ const cliCommands = [
     </div>
 
     <div class="flex flex-wrap items-center justify-between gap-3 px-1 text-cp-sm font-semibold text-cp-text-secondary">
-      <span>重新执行对应命令即可升级到当前发布版本。</span>
+      <span>重新执行对应命令即可升级到当前发布版本</span>
       <a
         href="https://learn.chatgpt.com/docs/codex/cli"
         target="_blank"

frontend/src/views/settings/components/client-version/components/ClientDesktopInstallPanel.vue

@@ -50,7 +50,7 @@ const desktopCommands = [
             在线安装与升级
           </h3>
           <p class="mt-1.5 mb-0 text-cp-sm leading-[1.45] font-semibold text-cp-text-secondary">
-            桌面应用通常会自动更新，也可以从官方渠道重新安装。
+            桌面应用通常会自动更新，也可以从官方渠道重新安装
           </p>
         </div>
         <a

frontend/src/views/settings/components/client-version/components/ClientDesktopOfflineDownloads.vue

@@ -75,7 +75,7 @@ function formatFileSize(value: number | null): string {
             </span>
           </div>
           <p class="mt-1.5 mb-0 text-cp-sm leading-[1.45] font-semibold text-cp-text-secondary">
-            后端实时校验 Microsoft CDN 直链；请选择与设备一致的架构。
+            后端实时校验 Microsoft CDN 直链，请选择与设备一致的架构
           </p>
         </div>
       </div>
@@ -102,7 +102,7 @@ function formatFileSize(value: number | null): string {
       class="m-0 rounded-cp bg-cp-error-container px-3 py-2.5 text-cp-sm leading-[1.45] font-bold text-cp-error-on-container"
       role="alert"
     >
-      {{ error }}，当前仍显示上次成功结果。
+      {{ error }}，当前仍显示上次成功结果
     </p>
 
     <div v-if="loading && !downloads" class="grid gap-3 sm:grid-cols-2" aria-label="正在提取离线安装包">

frontend/src/views/settings/composables/useBackupSettings.ts

@@ -137,7 +137,7 @@ export function useBackupSettings() {
         retentionCount: Number(schedule.retentionCount) || 0,
       })
       applySettings(data)
-      toast.success('调度配置已保存')
+      toast.success('备份计划已保存')
       return true
     }
     catch {

frontend/src/views/settings/composables/useSettingsForm.ts

@@ -23,7 +23,6 @@ export function useSettingsForm() {
   const savedRequestLocation = shallowRef<RequestLocation>()
   const form = reactive({
     openaiClientProfile: null as ClientProfileSelection | null,
-    disableFast: false,
     requestLocationEnabled: false,
     requestLocation: { country: '', region: '', city: '', timezone: '' },
     refreshMarginSeconds: null as number | null,
@@ -105,7 +104,6 @@ export function useSettingsForm() {
 
   function applySettings(data: Awaited<ReturnType<typeof getSettings>>) {
     savedRequestLocation.value = { ...data.requestLocation }
-    form.disableFast = data.disableFast
     form.requestLocationEnabled = data.requestLocationEnabled
     form.requestLocation = { ...data.requestLocation }
     form.refreshMarginSeconds = data.refreshMarginSeconds
@@ -189,7 +187,7 @@ export function useSettingsForm() {
       return
     const { refreshMarginSeconds, refreshConcurrency, maxConcurrentPerAccount, requestIntervalMs, rotationStrategy, maxWaitingPerKey, maxWaitingPerAccount, concurrencyWaitTimeoutSeconds, responsesMaxDecompressedBodyMiB, accountAutoFreezeThreshold, accountAutoFreezeWindowSeconds, accountAutoFreezeDurationSeconds } = form
     if (refreshMarginSeconds === null || refreshConcurrency === null || maxConcurrentPerAccount === null || requestIntervalMs === null || !rotationStrategy || maxWaitingPerKey === null || maxWaitingPerAccount === null || concurrencyWaitTimeoutSeconds === null) {
-      toast.warning('请完整填写运行参数和调度策略')
+      toast.warning('请完整填写并发、队列、凭据刷新参数和调度策略')
       return
     }
     if (responsesMaxDecompressedBodyMiB === null || !Number.isInteger(responsesMaxDecompressedBodyMiB) || responsesMaxDecompressedBodyMiB < 1
@@ -199,7 +197,7 @@ export function useSettingsForm() {
     }
     if (![maxWaitingPerKey, maxWaitingPerAccount].every(value => Number.isInteger(value) && value >= 0 && value <= 1000)
       || !Number.isInteger(concurrencyWaitTimeoutSeconds) || concurrencyWaitTimeoutSeconds < 1 || concurrencyWaitTimeoutSeconds > 120) {
-      toast.warning('最大排队数应为 0～1000 的整数，最长排队时间应为 1～120 秒的整数')
+      toast.warning('队列容量应为 0～1000 的整数，排队超时应为 1～120 秒的整数')
       return
     }
     if (minCodexDesktopVersionError.value || minCodexCliVersionError.value) {
@@ -216,13 +214,13 @@ export function useSettingsForm() {
       return
     }
     if (accountAutoFreezeThreshold === null || accountAutoFreezeWindowSeconds === null || accountAutoFreezeDurationSeconds === null) {
-      toast.warning('请完整填写账号自动冻结参数')
+      toast.warning('请完整填写过载保护参数')
       return
     }
     if (!Number.isInteger(accountAutoFreezeThreshold) || accountAutoFreezeThreshold < 2 || accountAutoFreezeThreshold > 1000
       || !Number.isInteger(accountAutoFreezeWindowSeconds) || accountAutoFreezeWindowSeconds < 60 || accountAutoFreezeWindowSeconds > 3600
       || !Number.isInteger(accountAutoFreezeDurationSeconds) || accountAutoFreezeDurationSeconds < 300 || accountAutoFreezeDurationSeconds > 604800) {
-      toast.warning('自动冻结阈值应为 2～1000，统计窗口为 60～3600 秒，冻结时长为 300～604800 秒')
+      toast.warning('失败次数阈值应为 2～1000，统计窗口为 60～3600 秒，冷却时长为 300～604800 秒')
       return
     }
     const probeModel = form.accountAutoFreezeProbeModel.trim()
@@ -234,7 +232,6 @@ export function useSettingsForm() {
     await saveAction.run(async () => {
       const result = await updateSettings({
         openaiClientProfile,
-        disableFast: form.disableFast,
         requestLocationEnabled: form.requestLocationEnabled,
         requestLocation,
         modelMappings: mappingPayload(),

frontend/src/views/settings/constants.ts

@@ -10,12 +10,12 @@ export const rotationOptions = [
     description: '优先选择额度窗口更快重置的账号，适合在重置前消耗剩余额度',
   },
   {
-    label: '轮询',
+    label: '轮询调度',
     value: 'round_robin',
     description: '在可用候选账号间按顺序轮转，分配结果最可预测',
   },
   {
-    label: '粘滞',
+    label: '粘性调度',
     value: 'sticky',
     description: '优先复用最近使用的账号，直到不可用后再切换',
   },

frontend/src/views/settings/index.vue

@@ -1,38 +1,50 @@
 <script setup lang="ts">
 import { Save, Undo2 } from '@lucide/vue'
-import { computed, ref, watch } from 'vue'
+import { computed, reactive, shallowRef, watch } from 'vue'
 import { useRoute, useRouter } from 'vue-router'
 
 import BaseButton from '@/components/base/BaseButton.vue'
 import BaseCard from '@/components/base/BaseCard.vue'
-import BaseConfirmModal from '@/components/base/BaseConfirmModal.vue'
 import BaseIconButton from '@/components/base/BaseIconButton.vue'
 import BasePageHeader from '@/components/base/BasePageHeader.vue'
 import BaseSegmented from '@/components/base/BaseSegmented.vue'
+import BaseSelect from '@/components/base/BaseSelect.vue'
 import ClientProfileEditor from '@/components/client-profile/ClientProfileEditor.vue'
 
 import AccountAutoFreezeCard from './components/AccountAutoFreezeCard.vue'
-import AdminApiKeyCard from './components/AdminApiKeyCard.vue'
-import AdminPasswordCard from './components/AdminPasswordCard.vue'
 import SettingsBackupSection from './components/backup/SettingsBackupSection.vue'
-import ClientVersionSettings from './components/client-version/index.vue'
-import FastPolicyCard from './components/FastPolicyCard.vue'
 import ModelAliasesCard from './components/ModelAliasesCard.vue'
 import RequestLocationCard from './components/RequestLocationCard.vue'
 import RequestQueueCard from './components/RequestQueueCard.vue'
-import ResponseBodyLimitCard from './components/ResponseBodyLimitCard.vue'
 import RotationStrategyCard from './components/RotationStrategyCard.vue'
 import RuntimeSettingsCard from './components/RuntimeSettingsCard.vue'
+import SettingsAccessSection from './components/SettingsAccessSection.vue'
 import TokenRefreshCard from './components/TokenRefreshCard.vue'
-import { useAdminApiKey } from './composables/useAdminApiKey'
 import { useSettingsForm } from './composables/useSettingsForm'
 import { rotationOptions } from './constants'
 import PricingSection from './pricing/index.vue'
 
 const route = useRoute()
 const router = useRouter()
-const pricingVisited = ref(false)
-const section = computed(() => route.name === 'settings-pricing' ? 'pricing' : route.name === 'settings-backup' ? 'backup' : 'runtime')
+const sectionOptions = [
+  { label: '网关调度', value: 'runtime' },
+  { label: '上游配置', value: 'upstream' },
+  { label: '模型定价', value: 'pricing' },
+  { label: '安全与访问', value: 'access' },
+  { label: '数据备份', value: 'backup' },
+]
+const section = computed(() => {
+  switch (route.name) {
+    case 'settings-upstream': return 'upstream'
+    case 'settings-access': return 'access'
+    case 'settings-pricing': return 'pricing'
+    case 'settings-backup': return 'backup'
+    default: return 'runtime'
+  }
+})
+const isBasicSection = computed(() => section.value !== 'pricing' && section.value !== 'backup')
+const visited = reactive(new Set<string>())
+const settingsVisited = shallowRef(false)
 
 function switchSection(value: string): void {
   void router.push(value === 'runtime' ? '/settings' : `/settings/${value}`)
@@ -67,65 +79,51 @@ const {
   loadSettings,
 } = useSettingsForm()
 
-const {
-  loading: adminKeyLoading,
-  regenerating: adminKeyRegenerating,
-  deleting: adminKeyDeleting,
-  showDeleteModal: showDeleteAdminKeyModal,
-  generatedKey: generatedAdminApiKey,
-  status: adminApiKeyStatus,
-  regenerate: handleRegenerateAdminApiKey,
-  remove: handleDeleteAdminApiKey,
-  copyGeneratedKey: copyAdminApiKey,
-  loadStatus: loadAdminApiKeyStatus,
-} = useAdminApiKey()
+const disabled = computed(() => saving.value || loading.value || !!error.value)
 
-const runtimeVisited = ref(false)
 watch(section, (value) => {
-  if (value === 'pricing') {
-    pricingVisited.value = true
-  }
-  else if (!runtimeVisited.value) {
-    runtimeVisited.value = true
+  visited.add(value)
+  if (isBasicSection.value && !settingsVisited.value) {
+    settingsVisited.value = true
     void loadSettings()
-    void loadAdminApiKeyStatus()
   }
 }, { immediate: true })
 </script>
 
 <template>
   <div class="w-full" :class="section === 'pricing' ? 'flex h-[calc(100dvh-2rem)] flex-none! flex-col min-[961px]:h-[calc(100dvh-3rem)]' : undefined">
-    <BasePageHeader title="系统设置" description="管理运行参数、模型定价、管理员凭据与备份配置" />
+    <BasePageHeader title="系统设置" description="管理网关调度、上游配置、模型定价、安全访问与数据备份" />
 
     <div class="mt-4 flex min-h-cp-control shrink-0 flex-wrap items-center justify-between gap-3">
       <BaseSegmented
         :model-value="section"
         label="设置分区"
-        class="bg-(--cp-input-bg)!"
-        :options="[
-          { label: '运行设置', value: 'runtime' },
-          { label: '模型定价', value: 'pricing' },
-          { label: '备份', value: 'backup' },
-        ]"
+        class="hidden! bg-(--cp-input-bg)! sm:inline-grid!"
+        :options="sectionOptions"
         @update:model-value="switchSection"
       />
-      <div v-if="section === 'runtime'" class="flex items-center justify-end gap-2">
+      <BaseSelect
+        :model-value="section"
+        aria-label="设置分区"
+        class="w-full sm:hidden"
+        :options="sectionOptions"
+        @update:model-value="switchSection"
+      />
+      <div v-if="isBasicSection || hasChanges" class="ml-auto flex items-center justify-end gap-2">
         <span v-if="hasChanges" class="mr-1 size-1.5 shrink-0 rounded-full bg-cp-warning" aria-hidden="true" />
-        <BaseIconButton v-if="hasChanges" label="撤销更改" variant="filled" :disabled="saving || loading" @click="resetSettings">
+        <BaseIconButton v-if="hasChanges" label="撤销全部基础设置更改" variant="filled" :disabled="saving || loading" @click="resetSettings">
           <Undo2 class="size-4" />
         </BaseIconButton>
         <BaseButton variant="primary" :loading="saving" :disabled="loading || !hasChanges || !!error" @click="saveSettings">
           <template #icon>
             <Save class="size-4" />
           </template>
-          {{ saving ? '保存中...' : '保存' }}
+          {{ saving ? '保存中...' : '保存基础设置' }}
         </BaseButton>
-        <span class="sr-only" role="status">{{ hasChanges ? '有未保存更改' : '' }}</span>
       </div>
     </div>
 
-    <!-- 切换页签只隐藏内容，保留运行设置和备份配置的未保存草稿。 -->
-    <div v-if="runtimeVisited" v-show="section === 'runtime'" class="mt-5 grid w-full gap-5">
+    <div v-if="settingsVisited" v-show="isBasicSection" class="mt-5 grid w-full gap-5">
       <div v-if="error" role="alert" class="flex flex-wrap items-center justify-between gap-3 rounded-cp-card bg-cp-error-container p-5 text-cp-error-on-container">
         <p class="m-0 text-cp">
           设置加载失败：{{ error }}
@@ -138,81 +136,59 @@ watch(section, (value) => {
         正在加载设置…
       </div>
 
-      <AdminPasswordCard />
-      <AdminApiKeyCard
-        :status="adminApiKeyStatus"
-        :loading="adminKeyLoading"
-        :regenerating="adminKeyRegenerating"
-        :deleting="adminKeyDeleting"
-        :generated-key="generatedAdminApiKey"
-        @regenerate="handleRegenerateAdminApiKey"
-        @request-delete="showDeleteAdminKeyModal = true"
-        @copy="copyAdminApiKey"
+      <SettingsAccessSection
+        v-if="visited.has('access')"
+        v-show="section === 'access'"
+        v-model:min-codex-desktop-version="form.minCodexDesktopVersion"
+        v-model:min-codex-cli-version="form.minCodexCliVersion"
+        v-model:responses-max-decompressed-body-mi-b="responsesMaxDecompressedBodyMiBValue"
+        :disabled="disabled"
+        :loading="loading"
+        :desktop-error="minCodexDesktopVersionError"
+        :cli-error="minCodexCliVersionError"
       />
 
-      <fieldset :disabled="saving || loading || !!error" class="m-0 grid min-w-0 gap-5 border-0 p-0" aria-label="运行设置">
-        <RuntimeSettingsCard
-          v-model:max-concurrent-per-account="maxConcurrentPerAccountValue"
-          v-model:request-interval-ms="requestIntervalMsValue"
-        />
-        <TokenRefreshCard v-model:refresh-margin-seconds="refreshMarginSecondsValue" v-model:refresh-concurrency="refreshConcurrencyValue" />
-        <ResponseBodyLimitCard v-model="responsesMaxDecompressedBodyMiBValue" />
-        <RequestQueueCard
-          v-model:max-waiting-per-key="maxWaitingPerKeyValue"
-          v-model:max-waiting-per-account="maxWaitingPerAccountValue"
-          v-model:concurrency-wait-timeout-seconds="concurrencyWaitTimeoutSecondsValue"
-        />
-        <FastPolicyCard v-model="form.disableFast" :disabled="saving || loading || !!error" />
+      <fieldset v-show="section !== 'access'" :disabled="disabled" class="m-0 grid min-w-0 gap-5 border-0 p-0" aria-label="基础设置">
+        <template v-if="section === 'runtime'">
+          <RuntimeSettingsCard
+            v-model:max-concurrent-per-account="maxConcurrentPerAccountValue"
+            v-model:request-interval-ms="requestIntervalMsValue"
+          />
+          <RotationStrategyCard v-model="form.rotationStrategy" :options="rotationOptions" />
+          <RequestQueueCard
+            v-model:max-waiting-per-key="maxWaitingPerKeyValue"
+            v-model:max-waiting-per-account="maxWaitingPerAccountValue"
+            v-model:concurrency-wait-timeout-seconds="concurrencyWaitTimeoutSecondsValue"
+          />
+          <AccountAutoFreezeCard
+            v-model:enabled="form.accountAutoFreezeEnabled"
+            v-model:threshold="accountAutoFreezeThresholdValue"
+            v-model:window-seconds="accountAutoFreezeWindowSecondsValue"
+            v-model:duration-seconds="accountAutoFreezeDurationSecondsValue"
+            v-model:probe-enabled="form.accountAutoFreezeProbeEnabled"
+            v-model:probe-model="form.accountAutoFreezeProbeModel"
+            v-model:adaptive-concurrency="form.accountAutoFreezeAdaptiveConcurrency"
+          />
+        </template>
 
-        <RequestLocationCard v-model="form.requestLocation" v-model:enabled="form.requestLocationEnabled" :disabled="saving || loading || !!error" />
-        <AccountAutoFreezeCard
-          v-model:enabled="form.accountAutoFreezeEnabled"
-          v-model:threshold="accountAutoFreezeThresholdValue"
-          v-model:window-seconds="accountAutoFreezeWindowSecondsValue"
-          v-model:duration-seconds="accountAutoFreezeDurationSecondsValue"
-          v-model:probe-enabled="form.accountAutoFreezeProbeEnabled"
-          v-model:probe-model="form.accountAutoFreezeProbeModel"
-          v-model:adaptive-concurrency="form.accountAutoFreezeAdaptiveConcurrency"
-        />
-
-        <BaseCard title="上游身份">
-          <ClientProfileEditor v-if="form.openaiClientProfile" v-model="form.openaiClientProfile" :disabled="saving || loading || !!error" class="max-w-6xl" />
-        </BaseCard>
-
-        <ClientVersionSettings
-          v-model:min-codex-desktop-version="form.minCodexDesktopVersion"
-          v-model:min-codex-cli-version="form.minCodexCliVersion"
-          :loading="loading"
-          :desktop-error="minCodexDesktopVersionError"
-          :cli-error="minCodexCliVersionError"
-        />
-
-        <ModelAliasesCard
-          :mappings="mappings"
-          :loading="loading"
-          @add-mapping="addMapping"
-          @update-mapping="updateMapping"
-          @remove-mapping="removeMapping"
-        />
-        <RotationStrategyCard v-model="form.rotationStrategy" :options="rotationOptions" />
+        <div v-if="visited.has('upstream')" v-show="section === 'upstream'" class="grid min-w-0 gap-5">
+          <TokenRefreshCard v-model:refresh-margin-seconds="refreshMarginSecondsValue" v-model:refresh-concurrency="refreshConcurrencyValue" />
+          <BaseCard title="客户端身份" description="配置网关向上游声明的客户端类型、版本与请求头">
+            <ClientProfileEditor v-if="form.openaiClientProfile" v-model="form.openaiClientProfile" :active="section === 'upstream'" :disabled="disabled" class="max-w-6xl" />
+          </BaseCard>
+          <RequestLocationCard v-model="form.requestLocation" v-model:enabled="form.requestLocationEnabled" :disabled="disabled" />
+          <ModelAliasesCard
+            :mappings="mappings"
+            :loading="loading"
+            @add-mapping="addMapping"
+            @update-mapping="updateMapping"
+            @remove-mapping="removeMapping"
+          />
+        </div>
       </fieldset>
     </div>
 
-    <SettingsBackupSection v-show="section === 'backup'" class="mt-5" :active="section === 'backup'" />
-    <PricingSection v-if="pricingVisited" v-show="section === 'pricing'" class="mt-5 min-h-0 flex-1" />
-
-    <BaseConfirmModal
-      v-model="showDeleteAdminKeyModal"
-      title="删除管理员 API Key"
-      description="删除后外部系统将无法继续使用该 Key 调用管理接口"
-      destructive
-      confirm-text="确认删除"
-      :loading="adminKeyDeleting"
-      @confirm="handleDeleteAdminApiKey"
-    >
-      <p class="m-0">
-        确定要删除当前管理员 API Key 吗？此操作会立即生效
-      </p>
-    </BaseConfirmModal>
+    <SettingsBackupSection v-if="visited.has('backup')" v-show="section === 'backup'" class="mt-5" :active="section === 'backup'" />
+    <PricingSection v-if="visited.has('pricing')" v-show="section === 'pricing'" class="mt-5 min-h-0 flex-1" />
   </div>
 </template>

frontend/src/views/settings/pricing/PricingBatchModal.vue

@@ -6,6 +6,7 @@ import BaseButton from '@/components/base/BaseButton.vue'
 import BaseFormItem from '@/components/base/BaseForm/FormItem.vue'
 import BaseInput from '@/components/base/BaseInput.vue'
 import BaseModal from '@/components/base/BaseModal/index.vue'
+import BaseScrollbar from '@/components/base/BaseScrollbar.vue'
 import { multiplierText, parseMultiplier } from './model'
 
 defineProps<{ rows: PricingRow[], reset: boolean, saving: boolean }>()
@@ -20,26 +21,28 @@ watch(open, (value) => {
 </script>
 
 <template>
-  <BaseModal v-model="open" :title="reset ? '清除人工覆盖' : '批量设置倍率'" :description="`仅修改下列 ${rows.length} 个模型；不重算历史账单`" :dismissible="!saving" tone="warning">
+  <BaseModal v-model="open" :title="reset ? '清除人工覆盖' : '批量设置倍率'" :description="`仅修改下列 ${rows.length} 个模型，不重算历史账单`" :dismissible="!saving" tone="warning">
     <div class="grid gap-4">
       <p v-if="reset" class="m-0 text-cp leading-relaxed text-cp-text-secondary">
-        将移除选中模型的人工单价和倍率，回到同步价目或内置价目。仅有人工价格的模型将恢复为未配置。
+        恢复来源或内置价格，无可用价格时设为未配置
       </p>
-      <BaseFormItem v-else label="目标倍率" description="设置为此倍率，不与已有倍率相乘；0～100，最多四位小数">
+      <BaseFormItem v-else label="目标倍率" description="设置为此倍率，不与已有倍率相乘，0～100，最多四位小数">
         <BaseInput v-model="multiplier" aria-label="目标倍率" inputmode="decimal" :disabled="saving">
           <template #suffix>
             ×
           </template>
         </BaseInput>
       </BaseFormItem>
-      <ul class="m-0 max-h-64 list-none overflow-auto rounded-cp bg-cp-fill-quaternary p-3 text-cp-sm">
-        <li v-for="row in rows" :key="row.model" class="flex items-start justify-between gap-3 py-2">
-          <span class="min-w-0 break-all font-mono">{{ row.model }}</span>
-          <span class="shrink-0 font-mono text-cp-text-secondary">{{ multiplierText(row.effective.multiplierBps) }} → {{ reset ? '1×' : bps === undefined ? '—' : multiplierText(bps) }}</span>
-        </li>
-      </ul>
+      <BaseScrollbar max-height="16rem" class="rounded-cp bg-cp-fill-quaternary">
+        <ul class="m-0 list-none p-3 text-cp-sm">
+          <li v-for="row in rows" :key="row.model" class="flex items-start justify-between gap-3 py-2">
+            <span class="min-w-0 break-all font-mono">{{ row.model }}</span>
+            <span class="shrink-0 font-mono text-cp-text-secondary">{{ multiplierText(row.effective.multiplierBps) }} → {{ reset ? '1×' : bps === undefined ? '—' : multiplierText(bps) }}</span>
+          </li>
+        </ul>
+      </BaseScrollbar>
       <p v-if="!reset && bps === undefined" role="alert" class="m-0 text-cp-sm text-cp-error">
-        请输入合法倍率。
+        请输入合法倍率
       </p>
     </div>
     <template #footer>

frontend/src/views/settings/pricing/PricingEditor.vue

@@ -67,22 +67,22 @@ function updatePrice(field: keyof TokenPrices, value: string) {
 }
 function submit() {
   if (!model.value || new TextEncoder().encode(model.value).length > 128 || /\s/.test(model.value) || [...model.value].some(char => char.charCodeAt(0) < 32 || char.charCodeAt(0) === 127)) {
-    validation.value = '模型 ID 必须为 1～128 字节，不含空白或控制字符。'
+    validation.value = '模型 ID 必须为 1～128 字节，不含空白或控制字符'
     return
   }
   if (bps.value === undefined) {
-    validation.value = '倍率范围为 0～100，最多四位小数。'
+    validation.value = '倍率范围为 0～100，最多四位小数'
     return
   }
   for (const [band, rates] of Object.entries(draft.value)) {
     if (Object.values(rates).some(price => !validPrice(price))) {
-      validation.value = '请完整填写该档位的四项单价：0～1000000，最多四位小数。'
+      validation.value = '请完整填写该档位的四项单价：0～1000000，最多四位小数'
       selectedBand.value = band as PriceBand
       return
     }
   }
   if (!props.row?.base.bands.standard && !draft.value.standard) {
-    validation.value = '新模型需要配置标准档单价。'
+    validation.value = '新模型需要配置标准档单价'
     selectedBand.value = 'standard'
     return
   }
@@ -141,13 +141,13 @@ function submit() {
         </template>
         <div :id="helpId" class="grid w-72 gap-2 p-3 text-cp-xs leading-relaxed text-cp-text-secondary">
           <p class="m-0">
-            模型 ID 使用实际发送给上游的名称，不是客户端别名。
+            模型 ID 使用实际发送给上游的名称，不是客户端别名
           </p>
           <p class="m-0">
-            自定义档位需填写四项单价，0 表示免费；单价上限 1000000，倍率范围 0～100，最多四位小数。
+            自定义档位需填写四项单价，0 表示免费，单价上限 1000000，倍率范围 0～100，最多四位小数
           </p>
           <p class="m-0">
-            仅调整本地估算与金额限额，不代表订阅实际扣费。保存后仅新请求生效，历史账单不变。
+            仅调整本地估算与金额限额，不代表订阅实际扣费，保存后仅新请求生效，历史账单不变
           </p>
         </div>
       </BasePopover>

frontend/src/views/settings/pricing/PricingSyncModal.vue

@@ -1,16 +1,18 @@
 <script setup lang="ts">
 import type { PricingCatalog, PricingSyncPreview } from '@/api'
-import { computed } from 'vue'
+import { computed, shallowRef, watch } from 'vue'
 import BaseButton from '@/components/base/BaseButton.vue'
+import BaseCheckbox from '@/components/base/BaseCheckbox.vue'
 import BaseModal from '@/components/base/BaseModal/index.vue'
 import BaseScrollbar from '@/components/base/BaseScrollbar.vue'
 import BaseTag from '@/components/base/BaseTag.vue'
 import { bands, effectivePrice, priceFields } from './model'
 
 const props = defineProps<{ preview?: PricingSyncPreview, catalog: PricingCatalog, saving: boolean }>()
-defineEmits<{ close: [], confirm: [] }>()
+const emit = defineEmits<{ close: [], confirm: [models: Record<string, string[]>] }>()
+const selected = shallowRef(new Set<string>())
 const changes = computed(() => {
-  const items: { id: string, custom: boolean, details: { label: string, before: string, after: string }[] }[] = []
+  const items: { id: string, provider: string, model: string, custom: boolean, details: { label: string, before: string, after: string }[] }[] = []
   const providers = new Set([...Object.keys(props.catalog.synced), ...Object.keys(props.preview?.prices ?? {})])
   for (const provider of providers) {
     const previous = props.catalog.synced[provider] ?? {}
@@ -32,37 +34,84 @@ const changes = computed(() => {
           }
         }
       }
-      items.push({ id: `${provider}/${model}`, details, custom: !!props.catalog.overrides[provider]?.[model] })
+      items.push({ id: `${provider}/${model}`, provider, model, details, custom: !!props.catalog.overrides[provider]?.[model] })
     }
   }
   return items
 })
+const selectedChanges = computed(() => changes.value.filter(item => selected.value.has(item.id)))
+const allSelected = computed(() => changes.value.length > 0 && selectedChanges.value.length === changes.value.length)
+const partiallySelected = computed(() => selectedChanges.value.length > 0 && !allSelected.value)
+
+watch(() => props.preview, () => {
+  selected.value = new Set(props.preview ? changes.value.map(item => item.id) : [])
+}, { immediate: true })
+
+function toggle(id: string, checked: boolean) {
+  const next = new Set(selected.value)
+  if (checked)
+    next.add(id)
+  else
+    next.delete(id)
+  selected.value = next
+}
+
+function confirm() {
+  if (props.saving || !selectedChanges.value.length)
+    return
+  const models: Record<string, string[]> = {}
+  for (const item of selectedChanges.value)
+    (models[item.provider] ??= []).push(item.model)
+  emit('confirm', models)
+}
 </script>
 
 <template>
-  <BaseModal :model-value="!!preview" title="确认同步来源价目" description="models.dev · USD / 1M Tokens" size="lg" :dismissible="!saving" @update:model-value="!$event && $emit('close')">
+  <BaseModal :model-value="!!preview" title="确认同步来源价目" description="USD / 1M Tokens" size="lg" :dismissible="!saving" @update:model-value="!$event && $emit('close')">
     <div class="grid gap-4">
       <p class="m-0 text-cp text-cp-text-secondary">
-        {{ changes.length }} 个模型的来源价格将变化。人工单价和倍率保留；未覆盖的档位会采用新的来源价格。models.dev 是社区价目，不代表订阅实际扣费。
+        仅同步所选模型的来源价格，保留自定义单价和倍率
       </p>
-      <div class="max-h-80 overflow-auto rounded-cp bg-cp-fill-quaternary p-4">
-        <div v-for="item in changes" :key="item.id" class="grid gap-2 py-3 text-cp-sm">
-          <div class="min-w-0">
-            <div class="break-all font-mono">
-              {{ item.id }}
-            </div><span v-if="item.custom" class="text-cp-xs text-cp-primary-text">保留人工覆盖</span>
-          </div>
-          <div v-for="detail in item.details" :key="detail.label" class="flex flex-wrap items-center justify-between gap-x-4 gap-y-1 text-cp-xs">
-            <span class="text-cp-text-secondary">{{ detail.label }}</span>
-            <span class="flex items-center gap-2 font-mono tabular-nums">
-              <span class="text-cp-text-tertiary">{{ detail.before }}</span><span aria-hidden="true">→</span><span class="text-cp-primary-text">{{ detail.after }}</span>
-            </span>
-          </div>
-        </div>
-        <p v-if="!changes.length" class="m-0 text-cp-sm text-cp-text-secondary">
-          来源价格没有变化。
-        </p>
+      <div v-if="changes.length" class="flex items-center justify-between gap-3">
+        <BaseCheckbox
+          :model-value="allSelected"
+          :indeterminate="partiallySelected"
+          label="全选"
+          show-label
+          :disabled="saving"
+          @update:model-value="selected = new Set($event ? changes.map(item => item.id) : [])"
+        />
+        <span class="text-cp-sm text-cp-text-tertiary" aria-live="polite">已选 {{ selectedChanges.length }} / {{ changes.length }}</span>
       </div>
+      <BaseScrollbar max-height="20rem" class="rounded-cp bg-cp-fill-quaternary">
+        <div class="p-4">
+          <div v-for="item in changes" :key="item.id" class="grid gap-2 py-3 text-cp-sm">
+            <BaseCheckbox
+              :model-value="selected.has(item.id)"
+              :label="item.id"
+              show-label
+              :disabled="saving"
+              @update:model-value="toggle(item.id, $event)"
+            >
+              <template #label>
+                <span class="break-all font-mono text-cp-sm font-normal">{{ item.id }}</span>
+              </template>
+            </BaseCheckbox>
+            <span v-if="item.custom" class="pl-6.5 text-cp-xs text-cp-primary-text">保留人工覆盖</span>
+            <div v-if="item.details.length" class="ml-6.5 grid gap-1">
+              <div v-for="detail in item.details" :key="detail.label" class="grid grid-cols-2 items-center gap-x-4 rounded-cp-sm bg-cp-bg-container px-2 py-1.5 text-cp-xs">
+                <span class="text-cp-text-secondary">{{ detail.label }}</span>
+                <span class="flex items-center gap-2 font-mono tabular-nums">
+                  <span class="text-cp-text-tertiary">{{ detail.before }}</span><span aria-hidden="true">→</span><span class="text-cp-primary-text">{{ detail.after }}</span>
+                </span>
+              </div>
+            </div>
+          </div>
+          <p v-if="!changes.length" class="m-0 text-cp-sm text-cp-text-secondary">
+            来源价格没有变化
+          </p>
+        </div>
+      </BaseScrollbar>
       <details v-if="preview?.skipped.length" class="text-cp-sm text-cp-text-secondary">
         <summary class="cursor-pointer">
           跳过 {{ preview.skipped.length }} 个缺少完整价格或计价方式不匹配的模型
@@ -81,7 +130,7 @@ const changes = computed(() => {
     <template #footer>
       <BaseButton :disabled="saving" @click="$emit('close')">
         取消
-      </BaseButton><BaseButton variant="primary" :loading="saving" @click="$emit('confirm')">
+      </BaseButton><BaseButton variant="primary" :loading="saving" :disabled="!selectedChanges.length" @click="confirm">
         确认同步
       </BaseButton>
     </template>

frontend/src/views/settings/pricing/PricingTable.vue

@@ -1,13 +1,15 @@
 <script setup lang="ts">
 import type { PricingRow } from './model'
 import type { BaseTableColumn } from '@/components/base/BaseTable/columns'
+import { Pencil, Trash2 } from '@lucide/vue'
 import BaseCheckbox from '@/components/base/BaseCheckbox.vue'
+import BaseIconButton from '@/components/base/BaseIconButton.vue'
 import BaseTable from '@/components/base/BaseTable/index.vue'
 import { effectivePrice, multiplierText, sourceLabels } from './model'
 import PricingUnit from './PricingUnit.vue'
 
 defineProps<{ rows: PricingRow[], selected: string[], loading: boolean, disabled: boolean }>()
-defineEmits<{ toggle: [model: string, checked: boolean], togglePage: [checked: boolean], edit: [row: PricingRow] }>()
+defineEmits<{ toggle: [model: string, checked: boolean], togglePage: [checked: boolean], edit: [row: PricingRow], delete: [row: PricingRow] }>()
 const columns: BaseTableColumn<PricingRow>[] = [
   { key: 'select', kind: 'selection' },
   { key: 'model', label: '模型', kind: 'identity' },
@@ -15,7 +17,7 @@ const columns: BaseTableColumn<PricingRow>[] = [
   { key: 'output', label: '输出', kind: 'numeric' },
   { key: 'cacheRead', label: '缓存读取', kind: 'numeric' },
   { key: 'multiplier', label: '倍率', kind: 'numeric', align: 'center' },
-  { key: 'actions', label: '操作', kind: 'actions', align: 'center' },
+  { key: 'actions', label: '操作', kind: 'actions' },
 ]
 
 function summaryBand(row: PricingRow) {
@@ -68,14 +70,14 @@ function summaryBand(row: PricingRow) {
       <span class="font-mono tabular-nums">{{ multiplierText(row.effective.multiplierBps) }}</span>
     </template>
     <template #actions="{ row }">
-      <button
-        type="button"
-        class="inline-flex h-cp-control shrink-0 cursor-pointer touch-manipulation items-center justify-center rounded-cp-sm border-0 bg-transparent p-0 text-cp leading-none font-bold text-cp-text-secondary outline-none transition-colors enabled:hover:text-cp-primary-text focus-visible:ring-2 focus-visible:ring-cp-control-outline disabled:cursor-not-allowed disabled:text-cp-text-disabled motion-reduce:transition-none"
-        :disabled="disabled"
-        @click="$emit('edit', row)"
-      >
-        编辑
-      </button>
+      <div class="flex items-center gap-1">
+        <BaseIconButton size="sm" label="编辑价格" :disabled="disabled" @click="$emit('edit', row)">
+          <Pencil class="size-3.5 text-cp-link" aria-hidden="true" />
+        </BaseIconButton>
+        <BaseIconButton v-if="row.canDelete" size="sm" label="删除价目" :disabled="disabled" @click="$emit('delete', row)">
+          <Trash2 class="size-3.5 text-cp-error" aria-hidden="true" />
+        </BaseIconButton>
+      </div>
     </template>
   </BaseTable>
 </template>

frontend/src/views/settings/pricing/PricingToolbar.vue

@@ -71,7 +71,7 @@ const providerOptions = [
             </template>
             <div :id="syncInfoId" class="grid w-64 gap-2 p-3 text-cp-xs leading-relaxed text-cp-text-secondary">
               <p class="m-0">
-                从 models.dev 获取社区价目，保留人工单价与倍率。
+                从 models.dev 获取社区价目，保留人工单价与倍率
               </p>
               <p class="m-0 text-cp-text-tertiary">
                 最近同步：{{ syncTime }}

frontend/src/views/settings/pricing/PricingUnit.vue

@@ -26,10 +26,10 @@ const infoId = useId()
         单价单位：USD / 1M Tokens
       </p>
       <p class="m-0">
-        人工覆盖优先于同步价目，再继承内置价格。仅影响本地估算与金额限额。
+        人工覆盖优先于同步价目，再继承内置价格，仅影响本地估算与金额限额
       </p>
       <p class="m-0">
-        表中为标准档生效价；图像模型显示图像 Token 价格。划线为来源价，其他档位在编辑中查看。
+        表中为标准档生效价，图像模型显示图像 Token 价格，划线为来源价，其他档位在编辑中查看
       </p>
     </div>
   </BasePopover>