langtrace

Why we included this project

Once an LLM application moves past the prototype stage, debugging becomes a matter of knowing what actually happened inside each model call. Langtrace is a self-hostable observability tool built on OpenTelemetry, so traces, latency, token counts, and cost show up for providers like OpenAI, Anthropic, and Gemini, plus vector databases and frameworks such as LangChain. Small TypeScript or Python SDKs wrap your existing code, and because traces follow OTEL conventions they drop into the monitoring stack you already run instead of living in a proprietary silo. Running it with Docker keeps production data inside your own infrastructure, and the dashboard is where you go to trace a bad prompt back to its cause or compare what different providers charge.