Reference · Glossary
Langfuse
Last updated
An open-source LLM observability platform — self-hostable or managed — that traces calls, tracks cost/latency/token usage per request, and supports prompt versioning and evaluation scores layered on top of production traffic.
#When to use
Teams that want tracing/observability data to stay in their own infrastructure (self-hosted), or need prompt version history alongside traces.
#When not to
A quick local prototype with a handful of test calls — the setup overhead isn't worth it until you have real traffic to observe.
#Example
Tagging traces by feature flag or prompt version, then comparing average latency and cost per version after a change — without exporting logs to a spreadsheet by hand.