[bot] Mistral Batch API (`client.batch.jobs`) is not instrumented
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
- 45/100
- Issue 类型
- 功能
- 描述清晰度
- 基本清楚
- 活跃度
- 活跃
- 技术栈
- javascript, typescript
- 领域
- observability, tooling
调研方向
Start by reading js/src/instrumentation/plugins/mistral-interceptors.ts to see how existing Mistral channels are declared, then mistral-plugin.ts to see how they are subscribed. Compare js/src/wrappers/mistral.ts and js/src/vendor-sdk-types/mistral.ts to understand how the batch namespace is currently unwrapped and typed. Done means client.batch.jobs.create/get/list and result download produce spans consistent with the existing chat and embeddings instrumentation, with tests covering the batch lifecycle.
由索引模型根据 Issue 内容生成。
描述
Summary
Mistral's official TypeScript SDK exposes a documented Batch API (client.batch.jobs.create()/.get()/.list(), plus client.files.download() for results) for submitting large-scale async chat, embeddings, FIM, moderation, classification, OCR, and conversation jobs. This repo instruments every one of Mistral's synchronous execution surfaces but has zero coverage of the batch lifecycle — no channel, no plugin logic, no vendor type.
What's missing
js/src/instrumentation/plugins/mistral-plugin.ts—MistralPlugin.onEnable()/subscribeToMistralChannels()(lines 24-277) subscribes to exactly ten channels:chatComplete,chatStream,embeddingsCreate,classifiersModerate,classifiersModerateChat,classifiersClassify,classifiersClassifyChat,fimComplete,fimStream,agentsComplete,agentsStream. There is nobatch-related channel anywhere in the file.js/src/instrumentation/plugins/mistral-interceptors.ts— grepped forbatch/Batch(case-insensitive): zero matches. NomistralChannels.batch*exists for the plugin to subscribe to in the first place.js/src/vendor-sdk-types/mistral.ts—MistralClient(lines 222-229) is typed as{ chat?, fim?, agents?, embeddings?, classifiers?, [key: string]: unknown }. There is nobatchnamespace type at all (noMistralBatch,MistralBatchJob, etc.), so even an untyped passthrough isn't modeled.js/src/wrappers/mistral.ts—wrapMistral()/mistralProxy()(lines 85-108) proxies onlychat,fim,agents,embeddings,classifiers. A client'sbatchproperty is returned as-is via the proxy'sdefaultbranch (Reflect.get, line 104) — fully unwrapped, soclient.batch.jobs.create(...)produces no span of any kind: no submission span, no status/polling span, no per-request result spans.
This means a customer submitting client.batch.jobs.create({ endpoint: "/v1/chat/completions", inputFiles: [...] }) — a fully supported, documented Mistral API call — gets no trace in Braintrust whatsoever, while the same customer's synchronous chat.complete/embeddings.create calls are fully traced with input, output, and token metrics.
Braintrust docs status
not_found. https://www.braintrust.dev/docs/providers/mistral documents only synchronous spans: "Chat completion spans (mistral.chat.complete...)", "Embedding spans (mistral.embeddings.create)", "Fill-in-the-middle completion spans", "Audio transcription spans." The page's "All call types support sync and async variants" line refers to Python's asyncio client variants, not batch jobs. No batch/async-job tracing is mentioned anywhere on the page.
Upstream sources
- https://docs.mistral.ai/capabilities/batch/ — Batch API guide:
client.batch.jobs.create()acceptsinput_files(file-based, up to 1,000,000 requests) or inlinerequests(up to 10,000); supportsendpointvalues/v1/embeddings,/v1/chat/completions,/v1/fim/completions,/v1/moderations,/v1/chat/moderations,/v1/ocr,/v1/classifications,/v1/conversations,/v1/audio/transcriptions; status viaclient.batch.jobs.get()/.list(); results retrieved viaclient.files.download().
Braintrust docs sources checked
- https://www.braintrust.dev/docs/providers/mistral
- https://www.braintrust.dev/docs/guides/tracing/integrations
Local repo files inspected
js/src/instrumentation/plugins/mistral-plugin.ts—subscribeToMistralChannels()(lines 24-277), full enumerated channel listjs/src/instrumentation/plugins/mistral-interceptors.ts— grepped forbatch/Batch, zero matchesjs/src/vendor-sdk-types/mistral.ts—MistralClienttype (lines 222-229)js/src/wrappers/mistral.ts—mistralProxy()(lines 85-108), proxy coverage limited tochat/fim/agents/embeddings/classifiers
- 主要语言
- TypeScript
- 星标
- 29
- 派生
- 21
- 平均合并
- 1 天 15 小时
- 30 天内合并 PR
- 76
环境准备
- 提供 Dockerfile 或 Docker Compose 文件
- 没有 Pull Request 模板
- 没有贡献指南
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。
braintrustdata/braintrust-sdk-javascript 的其他 Issue
-
typescript
难度 3/5 1-2 天 新手友好度 45/100
braintrustdata/braintrust-sdk-javascript#2686 ·
维护者通常 1 天内回复
-
[bot] Cloudflare Workers AI `queueRequest` (batch) inference is explicitly excluded from tracing未关闭cloudflare typescript
难度 4/5 3-5 天 新手友好度 30/100
braintrustdata/braintrust-sdk-javascript#2684 ·
维护者通常 1 天内回复
-
High TypeScript checking cost from generated Zod 3 schemas可能已有人在做 @lforst 于 14 天前认领。 未关闭
难度 4/5 3-5 天 新手友好度 22/100
braintrustdata/braintrust-sdk-javascript#2624 · 1 条评论 · 已指派 1 人 ·
维护者通常 1 天内回复
-
[bot] Instrument Luma AI (17,324 weekly downloads)可能重新可做 关联的 PR 已关闭且未合并。 未关闭typescript
难度 4/5 3-5 天 新手友好度 55/100
braintrustdata/braintrust-sdk-javascript#2578 ·
维护者通常 1 天内回复
-
[bot] Instrument Letta Agent SDK (87,205 weekly downloads)可能重新可做 关联的 PR 已关闭且未合并。 未关闭typescript
难度 4/5 3-5 天 新手友好度 55/100
braintrustdata/braintrust-sdk-javascript#2577 ·
维护者通常 1 天内回复
查看 braintrustdata/braintrust-sdk-javascript 的全部 Issue
相似的 Issue
-
难度 2/5 1-3 小时 新手友好度 65/100
lichess-org/lila#22324 · 1 条评论 ·
维护者通常 1 天内回复
-
ep:WebGPU platform:web quantization
难度 2/5 1-3 小时 新手友好度 66/100
microsoft/onnxruntime#33752 ·
维护者通常 2 天内回复
-
[Skill Bug]: entra-agent-id: inaccurate blueprint permission inheritance and parent token sub claim未关闭bug entra-agent-id
难度 2/5 1-3 小时 新手友好度 78/100
microsoft/GitHub-Copilot-for-Azure#3467 · 1 条评论 ·
维护者通常 1 天内回复
-
fix(worker): keep normal prepare waiting states failure-free可能已有人在做 @LichKing-2234 今天认领。 未关闭agent:codex priority:P1
难度 2/5 1-3 小时 新手友好度 72/100
AgoraIO-Extensions/agent-infra#3745 ·
维护者通常 1 天内回复
-
The five-line work window can drop its oldest line: the per-line byte share ignores the JSON framing可能已有人在做 @Eapp1e 今天认领。 未关闭
难度 2/5 1-3 小时 新手友好度 84/100
makecindy/cindy#7973 · 1 条评论 ·
维护者通常 1 天内回复