Command code GOAT used via API: MiMo V2.6 Pro stream consistently fails at ~785s with
まだ誰も着手していません。
評価
- 難易度
- 4/5
- 見積もり時間
- 3〜5日
- 初心者へのやさしさ
- 35/100
- issue の種類
- バグ
- 明瞭さ
- おおむね明確
- 活発さ
- 活発
- 領域
- api, backend, observability
調査の方向性
Start by inspecting CommandCode trace b416c51c0eb32fde7cd20646bfd99625 and the Provider API route for the xiaomi/mimo-v2.6-pro request. Compare the 785.097-second termination with the gateway_stream_timeout and the COMPLETED record in CommandCode Studio. Done means identifying the timeout source and determining whether the failure status and upstream handling need correction.
索引モデルが issue の本文から書いたものです。
説明
Summary
I am using the CommandCode GOAT API through OpenCode in Zed. Long-running autonomous agent sessions occasionally fail when an individual xiaomi/mimo-v2.6-pro inference runs for approximately 13 minutes.
The latest failure produced:
Internal error: {"code":"gateway_stream_timeout","message":"Stream exceeded maximum duration before function timeout","origin":"gateway"}: {
"service": "session",
"errorName": "UnknownError"
}
The corresponding request in CommandCode Studio is:
Model: xiaomi/mimo-v2.6-pro
Duration: 785.097s
Input tokens: 0
Output tokens: 0
Cost: $0.0000
Status: COMPLETED
Trace ID: b416c51c0eb32fde7cd20646bfd99625
Dashboard timestamp: Sep 27, 2:54:10 PM (as displayed in Studio).
Expected Behavior
Continue streaming without ~13 minutes timeouts.
Actual Behavior
"Stream exceeded maximum duration before function timeout"
Steps to reproduce the issue
- Configure OpenCode in Zed to use the CommandCode API (Provider API?).
- Select
xiaomi/mimo-v2.6-pro. - Run a long autonomous coding-agent session. ZDR is not enabled.
- When an individual MiMo inference runs for approximately 13 minutes, the request terminates with:
Internal error: {"code":"gateway_stream_timeout","message":"Stream exceeded maximum duration before function timeout","origin":"gateway"}: {
"service": "session",
"errorName": "UnknownError"
}
- In CommandCode Studio, the corresponding request is recorded as:
Model: xiaomi/mimo-v2.6-pro
Duration: 785.097s
Input tokens: 0
Output tokens: 0
Cost: $0.0000
Status: COMPLETED
Trace ID: b416c51c0eb32fde7cd20646bfd99625
Other MiMo requests in the same session completed successfully, including requests lasting approximately 545s, 399s, and 317s.
The overall agent session was about 32m25s long. The failure appears tied to the duration of one individual streaming inference, not the total session duration.
Command Code Version
N/A
Operating System
Linux
Terminal/IDE
Zed + OpenCode via ACP
Shell
N/A — issue occurs through Zed/OpenCode
Session file (optional)
Not attached. The relevant CommandCode trace is:
b416c51c0eb32fde7cd20646bfd99625
Fix prompt (optional)
Please inspect trace b416c51c0eb32fde7cd20646bfd99625 and determine which gateway or upstream generated the timeout.
The request terminates at 785.097 seconds with:
gateway_stream_timeout
Stream exceeded maximum duration before function timeout
Please check whether the MiMo V2.6 Pro Provider API route has an approximately 785-second maximum streaming duration.
If this limit is imposed by an upstream gateway, please consider routing long MiMo requests through an upstream without this limit, or implementing transparent retry/failover.
Also check why CommandCode Studio records this timed-out request as COMPLETED with 0 input tokens, 0 output tokens, and $0 cost instead of marking it as failed.
Additional context
This has happened several times during one day on the same account - though only this one occurrence is captured.
The timing looks suspicious:
There is a recent Vercel AI Gateway report describing streaming requests being terminated at approximately 785–790 seconds with the same:
gateway_stream_timeout
Stream exceeded maximum duration before function timeout
https://community.vercel.com/t/ai-gateway-streaming-requests-terminate-at-785-790s-despite-a-1800s-function-duration/48725
I don't know whether this CommandCode route uses Vercel AI Gateway or whether this is an equivalent timeout elsewhere in the routing stack, but the timing and error payload appear nearly identical.
- 主要言語
- 言語のデータがありません
- スター
- 4k
- フォーク
- 357
- PR マージ指標
- 30日以内にマージされた PR はありません
環境構築
このプロジェクトには開発コンテナ、Dockerfile、コントリビューションガイドがありません。まず README を読み、一般的な手順ははじめてのコントリビューションガイドを参照してください。
はじめの一歩
- issue を最後まで読み、次にプロジェクトのコントリビューションガイドを読みます。
- 着手することを issue にコメントします — 二人が同じ作業をするのを防げます。
- リポジトリをフォークし、ブランチを切って変更します。
- issue 番号を参照したプルリクエストを送ります。
CommandCodeAI/command-code のほかの issue
-
難易度 2/5 1〜3時間 初心者へのやさしさ 68/100
CommandCodeAI/command-code#933 ·
-
難易度 2/5 1〜3時間 初心者へのやさしさ 68/100
CommandCodeAI/command-code#855 ·
-
難易度 2/5 1〜3時間 初心者へのやさしさ 78/100
CommandCodeAI/command-code#841 · コメント 1 件 ·
-
難易度 2/5 1〜3時間 初心者へのやさしさ 68/100
CommandCodeAI/command-code#655 · コメント 1 件 ·
-
難易度 2/5 1〜3時間 初心者へのやさしさ 68/100
CommandCodeAI/command-code#608 ·
CommandCodeAI/command-code の issue をすべて見る
似ている issue
-
API Bug
難易度 2/5 1〜3時間 初心者へのやさしさ 78/100
ProjectSidewalk/SidewalkWebpage#5556 ·
メンテナーはふだん 1 日以内に返信
-
backend::vllm diffusion multimodal
難易度 2/5 1〜3時間 初心者へのやさしさ 72/100
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 84/100
openai/openai-agents-python#5229 ·
メンテナーはふだん 1 日以内に返信
-
難易度 2/5 1〜3時間 初心者へのやさしさ 84/100
メンテナーはふだん 1 日以内に返信
-
priority/4/normal status/needs-triage type/bug/unconfirmed
難易度 2/5 1〜3時間 初心者へのやさしさ 78/100
authelia/authelia#13292 · コメント 1 件 ·
メンテナーはふだん 1 日以内に返信