mirror of
https://github.com/zyycn/codex-proxy-rs.git
synced 2026-11-06 16:06:08 +00:00
fix(openai): restrict capacity cooldown to explicit overload
Count only confirmed capacity rejections toward account cooldown, excluding generic server failures and unrecognized upstream errors. Classify streamed invalid_prompt failures as invalid requests, matching official Codex. Cover HTTP and WebSocket failures around the cooldown threshold while preserving retry boundaries, upstream error payloads, and Smart scoring. Update the capacity cooldown contract in the API and architecture docs. Validation: Rustfmt, strict workspace Clippy, and all backend suites with isolated PostgreSQL 18 / Redis 8 (--test-threads=4). Reference: openai/codex@7498521d28, codex-rs/codex-api/src/sse/responses.rs. Refs #177 Co-authored-by: Codex <noreply@openai.com>
This commit is contained in:
+3
-2
@@ -1153,8 +1153,9 @@ User-Agent 使用 `grok-shell/<版本> (<系统>; <架构>)`,其中 `arm64`
|
||||
|
||||
### 账号自动冻结
|
||||
|
||||
账号自动冻结(`accountAutoFreezeEnabled`)默认关闭。启用后,在统计窗口内按尝试累计容量类上游错误(`server_is_overloaded`
|
||||
等与 5xx 不可用),达到阈值后把该账号冻结为带恢复倒计时的 `rate_limited` 状态。`accountAutoFreezeThreshold`
|
||||
账号自动冻结(`accountAutoFreezeEnabled`)默认关闭。启用后,在统计窗口内按尝试累计明确的上游容量拒绝
|
||||
(`server_is_overloaded`、`slow_down` 或结构化错误中的明确过载提示),普通 5xx、`invalid_prompt` 和未识别的上游错误不计入。
|
||||
达到阈值后把该账号冻结为带恢复倒计时的 `rate_limited` 状态。`accountAutoFreezeThreshold`
|
||||
取值 2~1,000(默认 12,按普通请求的 attempt 计数,含请求内同账号重试,不含诊断探测与本地连接保护错误);`accountAutoFreezeWindowSeconds`
|
||||
取值 60~3,600(默认 600,随每次失败滑动顺延);`accountAutoFreezeDurationSeconds` 取值 300~604,800
|
||||
(默认 7,200,即 2 小时,探测失败后按该时长顺延)。`accountAutoFreezeProbeEnabled` 开启时恢复 worker
|
||||
|
||||
Reference in New Issue
Block a user