Langfuse v4.50.0
Langfuse v4.50.0 introduces evaluation enhancements including facet types, built-in protection, rule idle time, and 7-day health tracking, along with API metadata typing fixes.
原始内容为英文；当前页面提供中文导航与来源说明，具体事实请以原文为准。
为什么值得关注
Langfuse release v4.50.0 brings several updates focused on the evaluations system and API stability. Features added include facet types, built-in protection, and rule idle time for evaluations, along with 7-day execution health visibility on evaluator statuses. Evaluation setup reliability was improved by validating evaluator models before saving and checking evaluator model availability. On the API side, read-side metadata for traces, observations, and scores is now typed as objects. UI and web updates include displaying error IDs in tRPC error toasts for issue reporting, showing full metadata JSONPath values in the modern session header, displaying view owner info on hover, and labeling S3 blob storage exports as Amazon S3. Dependencies such as Next.js and Piscina were also upgraded.
可执行摘要
Langfuse updated to v4.50.0. Key changes include adding facet types, protection, and rule idle time to evals, showing seven-day execution health, validating evaluator models before saving, typing read-side metadata on traces/observations/scores as objects in the API, and dependency bumps.
- Agent 实用度
- 77/100
- 可信度
- 96%
- 机器格式
- JSON + Markdown
开发者应核对什么
- Verify client-side parsing for trace, observation, and score metadata to ensure compatibility with object typing on the read-side API.
- Review new evaluator configuration options such as facet types, built-in protection, and rule idle time when configuring evaluation pipelines.