Verified project record

langfuse/langfuse

Langfuse is an open source AI engineering platform for collaboratively developing, monitoring, evaluating, and debugging LLM calls. It provides observability, prompt management, datasets, and a playground through a Web GUI and API.

Project overview

The platform supports self-hosting through Docker and Kubernetes while providing automated instrumentation for widely used LLM frameworks.

Project type: Workflow · Evaluation & Observability
Use cases: Data Analysis
Deployment: Docker / Docker Compose
License: License pending

Best for

AI engineers, developers, and data teams who need to collaboratively develop, monitor, evaluate, and debug LLM calls and application logic.

Key capabilities

Instrument your application to ingest traces, tracking LLM calls and other relevant logic such as retrieval, embedding, or agent actions.
Supports LLM-as-a-judge, code evaluators, user feedback collection, manual labeling, and custom evaluation pipelines via APIs or SDKs.
Enables test sets and benchmarks for evaluating LLM applications to support pre-deployment testing and structured experiments.
A tool for testing and iterating on prompts and model configurations to shorten the feedback loop.
A public API accompanied by an OpenAPI spec, Postman collection, and typed SDKs to power bespoke LLMOps workflows.

Limitations and risks

The application relies on the ClickHouse open source database.
Self-hosted instances automatically report basic usage statistics to a centralized server by default.
Folders containing enterprise edition features are not MIT licensed.
Using the LLM-as-a-judge evaluation feature optionally requires an external LLM API.

Getting started

Clone the repository and run docker compose up to access the local Web GUI. From there, create a new project and API credentials.

Alternatives and comparisons

Provides declarative testing and automated evaluations for LLM prompts, agents, and RAGs.
A Python framework to add Input/Output Guards that validate LLM interactions and to generate structured data from LLMs.

Project comparisons

Evidence and sources

Open original GitHub project