Lunary

Lunary 是给 LLM 应用做的可观测与运营平台：把每次对话的请求、响应、token 与延迟记成可检索的追踪，提示词在后台做版本管理和 A/B，评测与用户反馈挂在同一条会话上，出了问题能顺着 trace 一路查到具体哪一步。支持自托管，数据不出自己的机器，对合规敏感的团队是加分项。文档里有 OpenAI、LangChain、LiteLLM 等框架的接入示例，按 Python 与 JS 两套 SDK 分开写。 🔗 文档: https://docs.lunary.ai/

Lunary is an observability and operations platform for LLM applications: every conversation's request, response, tokens and latency become a searchable trace, prompts get versioning and A/B in the backend, and evaluations and user feedback hang off the same thread so a problem can be followed step by step. It self-hosts, keeping data on your own machines, which matters for compliance-sensitive teams. Docs cover OpenAI, LangChain and LiteLLM integration, split into Python and JavaScript SDKs. ⭐ Notable: trace-level debugging, prompt versioning, self-hosting 🔗 Docs: https://docs.lunary.ai/