Feature request: Retry-After header on 429 responses for exhausted Token Plan quota (Open Platform API)
维护者通常 1 天内回复
还没有人认领这个 Issue。
评估
- 难度
- 4/5
- 预计耗时
- 3-5 天
- 新手友好度
- 15/100
- Issue 类型
- 功能
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
调研方向
The request concerns the Open Platform API server's 429 responses for error 2056, and the payload names no file, test or entry point in this repository. Confirm first whether the CLI code here produces or handles those responses at all, since the header would need to be set by the server. Done means a 429 for an exhausted Token Plan quota includes a Retry-After header whose value is the seconds until the window resets.
由索引模型根据 Issue 内容生成。
描述
Feature Request: include Retry-After header in 429 responses for exhausted Token Plan quota
(This is about the Open Platform API server, not the CLI itself — filing here since this is MiniMax's official public repository.)
Scenario
Agent applications run a primary/fallback model chain with MiniMax as the primary. When the Token Plan quota is exhausted, the 429 response (error 2056, "已达到 Token Plan 用量上限") currently carries no Retry-After header.
Problem
Downstream agent frameworks (e.g. OpenClaw) use Retry-After to classify the rate-limit window:
- With
Retry-After(long window): the client fails over to the fallback model immediately — zero retries. - Without
Retry-After: the client treats it as a transient throttle and keeps retrying the exhausted provider ~10 times over 2–3 minutes before giving up.
During the quota window, every user message pays this 2–3 minute delay, because each turn starts from the primary model again.
Request
Add a Retry-After header to 429 responses that indicate exhausted usage windows (error 2056 Token Plan quota, and similar), with the value set to the remaining seconds until the quota window resets.
Precedent
A comparable CN provider already returns Retry-After on the same class of error (quota exhausted for a billing window), and its clients fail over in milliseconds.
Impact
All agent applications using MiniMax as a primary model benefit: failover latency during quota windows drops from minutes to seconds. Low-cost change, standard HTTP 429 semantics.
No account identifiers, request IDs, or timestamps included — no additional information should be needed to evaluate this request.
- 主要语言
- TypeScript
- 星标
- 2.2k
- 派生
- 190
- 平均合并
- 1 小时 39 分钟
- 30 天内合并 PR
- 3
环境准备
这个项目没有提供开发容器、Dockerfile 或贡献指南，环境需要你自己搭建：先看它的 README，通用步骤见我们的新手贡献指南。
从这里开始
- 先读完整个 Issue，再读项目的贡献指南。
- 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
- Fork 仓库，在一个分支上完成修改。
- 提交 Pull Request，并在描述里引用这个 Issue 编号。
MiniMax-AI/cli 的其他 Issue
-
[Bug] 未登录时 mmx help 被认证检查拦截，无法查看公开文档链接可能重新可做 关联的 PR 已关闭且未合并。 未关闭
难度 2/5 1-3 小时 新手友好度 84/100
MiniMax-AI/cli#259 · 1 条评论 ·
维护者通常 1 天内回复
-
[Bug] --no-color 未去除请求状态栏的 ANSI 颜色可能重新可做 关联的 PR 已关闭且未合并。 未关闭
难度 2/5 1-3 小时 新手友好度 88/100
MiniMax-AI/cli#258 ·
维护者通常 1 天内回复
-
难度 3/5 1-2 天 新手友好度 63/100
MiniMax-AI/cli#278 ·
维护者通常 1 天内回复
-
SDK streaming iterators leave their response bodies open on early exit可能已有人在做 @dvd233 于 11 天前认领。 未关闭
难度 3/5 1-2 天 新手友好度 57/100
MiniMax-AI/cli#277 ·
维护者通常 1 天内回复
-
/compact command未关闭
难度 5/5 一周以上 新手友好度 25/100
MiniMax-AI/cli#275 ·
维护者通常 1 天内回复
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 78/100
维护者通常 1 天内回复
-
bug
难度 2/5 1-3 小时 新手友好度 68/100
jbacalso24/mimry#459 · 1 条评论 ·
维护者通常 1 天内回复
-
architecture help wanted hive/claimed-by-architect origin/architect stage/030-pr-in-progress tech-debt
难度 2/5 1-3 小时 新手友好度 78/100
kubestellar/console#24422 · 1 条评论 ·
维护者通常 1 天内回复
-
area: backends type: docs
难度 1/5 1-3 小时 新手友好度 83/100
schlessera/brain-kit#2383 ·
维护者通常 1 天内回复
-
agent duplicate providers tool triaged
难度 2/5 1-3 小时 新手友好度 74/100
维护者通常 1 天内回复