openlit

Why we included this project

Teams building LLM features usually end up with one tool for tracing, another for cost, a third for evals, and a separate place to manage prompts. OpenLIT bundles those jobs into one self-hosted platform that emits standard OpenTelemetry traces from Python, TypeScript, or Go services and shows them on a single dashboard. It connects to more than fifty LLM providers, vector databases, and agent frameworks, which makes it a reasonable fit for heterogeneous stacks rather than a single vendor's SDK. The rule engine and eleven automated evaluation types cover hallucination, toxicity, and faithfulness checks without a separate judging pipeline, and the built-in vault keeps API keys in one place. For a small team without platform engineers, that combination is a practical way to get observability, evaluations, and prompt management running together.