Phoenix is an open-source AI observability platform for experimenting with, evaluating, and troubleshooting LLM applications. It uses OpenTelemetry-based tracing to help developers monitor and improve AI features.
Project overview
Provides a vendor, language, and framework-agnostic approach to LLM observability through OpenTelemetry, allowing teams to instrument and benchmark applications without deep vendor lock-in.
AI engineers, developers, and data teams who need to evaluate, observe, and troubleshoot LLM applications.
Key capabilities
Trace an LLM application's runtime using OpenTelemetry-based instrumentation.
Leverage LLMs to benchmark application performance using response and retrieval evals.
Create versioned datasets of examples for experimentation, evaluation, and fine-tuning.
Track and evaluate changes to prompts, LLMs, and retrieval.
Provides a unified interface to Phoenix's capabilities via Model Context Protocol.
Limitations and risks
When deployed using the Azure template, the application serves plain HTTP and requires a TLS proxy for production use.
Getting started
Installation is considered easy because the application is installable via pip/conda. The first success path is running pip install arize-phoenix. Users can interact with it via a web GUI, CLI, or API. Telemetry is disabled by default.
Alternatives and comparisons
Provides an open source AI engineering platform with LLM observability, evaluations, prompt management, datasets, and a playground.
Focuses on declarative testing, automated evaluations, and vulnerability scanning for LLM prompts, agents, and RAGs.
README: Phoenix is an open-source AI observability platform designed for experimentation, evaluation, and troubleshooting. It provides:
README: Install Phoenix via `pip` or `conda` and have a fully functional Phoenix ```shell pip install arize-phoenix phoenix serve ```
README: Phoenix container images are available via [Docker Hub](https://hub.docker.com/r/arizephoenix/phoenix) and can be deployed using Docker or Kubernetes. Arize AI also provides cloud…
Release: arize-phoenix: v18.1.0
README: MCP server implementation for Arize Phoenix providing unified interface to Phoenix's capabilities