Arize Phoenix

Arize Phoenix is an open-source LLM observability platform that you can run locally to trace, visualize, and evaluate LLM application behavior. It provides deep visibility into what your AI application is doing without sending data to external services.

Key use cases include:

• Trace visualization — see the complete execution flow of every LLM call, retrieval, and tool use
• Retrieval quality analysis — evaluate how well your RAG pipeline retrieves relevant documents
• Embedding analysis — visualize and detect drift in embedding spaces
• LLM-as-judge evaluations — run automated quality assessments directly in the UI
• Production monitoring — detect quality degradation and anomalies in real-time
• Debugging — drill into specific traces to understand why outputs were wrong

Phoenix is used by ML engineers who want self-hosted observability without sending sensitive production data to third-party services. It is built on OpenTelemetry standards and auto-instruments LangChain, LlamaIndex, OpenAI, and other frameworks.

Also available as Arize Phoenix Cloud for teams who prefer a hosted solution.

1. Install Phoenix:

   pip install arize-phoenix

2. Launch the UI:

   python -m phoenix.server.main serve
   # Opens at http://localhost:6006

3. Instrument your application:

   from phoenix.otel import register
   tracer_provider = register(endpoint='http://localhost:6006/v1/traces')
   
   from openinference.instrumentation.openai import OpenAIInstrumentor
   OpenAIInstrumentor().instrument(tracer_provider=tracer_provider)

4. View traces and run evaluations in the Phoenix UI at localhost:6006.

Pricing: Phoenix is free and open source (Apache 2.0 + Elastic v2). Self-host at no cost. Phoenix Cloud offers free and paid tiers. The full Arize platform provides additional enterprise features.