Arize Phoenix
ActiveDescription
Phoenix is an open-source observability and evaluation tool for LLM and agent applications, supporting online tracing and offline diagnosis.
Key Features
- OpenTelemetry-based tracing for LLM application runtime observability
- LLM-powered evaluation for response and retrieval quality benchmarking
- Versioned datasets for experimentation, evaluation, and fine-tuning
- Prompt management with version control, tagging, and experimentation
- Playground to optimize prompts, compare models, and replay traced calls
- Built-in PXI agent for debugging traces and navigating Phoenix
Use Cases
π‘ Trace and debug LLM calls across LangChain, LlamaIndex, and OpenAI SDK
π‘ Evaluate RAG pipeline retrieval quality with built-in evals
π‘ Compare prompt versions and model variants systematically
π‘ Monitor production LLM performance and detect regressions
π‘ Manage prompt libraries with version control and A/B testing
Tags
Categories
Quick Start
pip install arize-phoenix β import phoenix as px β px.launch_app() β open http://localhost:6006 β instrument your LLM code with OpenTelemetry