Arize-ai/phoenix

Health v2 · maintenance only

Flags: no_license

Adoption not part of the score

11205 stars · 1080 forks observed · 2026-08-28

What it is AI-extracted, prompt v1, taxonomy v1, 2026-08-29, confidence not recorded

Arize Phoenix is an open-source AI observability and evaluation platform for tracing, debugging, and evaluating LLM applications. It supports OpenTelemetry-based tracing, LLM-as-a-judge evals, prompt iteration, datasets, and experiments, and can be self-hosted for free.

Use cases

• trace and debug LLM application runs
• evaluate LLM outputs with LLM-as-a-judge
• catch regressions in AI app quality
• iterate on prompts using production examples
• run experiments comparing prompt or model changes
• self-host AI observability with data staying on-prem
• instrument LangChain or LlamaIndex apps with OpenTelemetry

When to choose

• you need end-to-end tracing for LLM apps built with LangChain, LlamaIndex, OpenAI, or Anthropic
• you want free, fully-featured self-hosted LLM observability with no data leaving your infrastructure
• you need systematic evaluation of RAG or agent outputs including LLM-as-a-judge scoring
• you want to compare prompt and model changes with datasets and experiments

When to avoid

• you only need simple application performance monitoring without LLM-specific features
• you want a fully managed enterprise observability platform with vendor support (consider Arize AX)
• your stack has no LLM or AI components

Facets

application · maturity active

monitoring tracing llm-inference rag prompt-engineering agent-framework benchmarking large-language-models monitoring developer-tools machine-learning python self-hosted cli llmops opentelemetry llm-evaluation llm-as-a-judge openinference prompt-management experiments tracing ai-agents docker kubernetes web-server

Member repositories

For agents

markdown · JSON · MCP: product_card(name="Arize-ai/phoenix")