Skip to main content

Phoenix Integration

Phoenix is Arize AI's open-source ML observability platform for LLM applications.

Overview​

The Phoenix integration enables LangGuard to:

  • Import traces from Phoenix
  • Sync spans and operation data
  • Track embeddings and retrievals
  • Apply governance policies

Prerequisites​

  • Phoenix instance (self-hosted or cloud)
  • API access to Phoenix
  • Traces logged to Phoenix

Phoenix Deployment Options​

Arize Cloud​

Use Arize's hosted Phoenix:

PHOENIX_API_KEY=your-arize-api-key
PHOENIX_HOST=https://app.phoenix.arize.com

Self-Hosted​

Run Phoenix locally or on your infrastructure:

# Start Phoenix
pip install arize-phoenix
python -m phoenix.server.main serve

# Configure LangGuard
PHOENIX_HOST=http://localhost:6006

Setup​

Add Integration​

  1. Navigate to Settings > Integrations in LangGuard
  2. Click Add Integration
  3. Select Phoenix
  4. Enter:
    • Name: Friendly name
    • Host URL: Phoenix server URL
    • API Key: (if required)
  5. Click Test Connection
  6. Configure sync settings
  7. Click Save

What Gets Synced​

Phoenix EntityLangGuard Mapping
TracesTraces
SpansSpan details
DocumentsRetrieved contexts
EmbeddingsVector metadata

Features​

Embedding Analysis​

Phoenix excels at embedding visualization:

  • Document retrieval quality
  • Embedding drift detection
  • Cluster analysis

RAG Evaluation​

Track retrieval-augmented generation:

  • Retrieval relevance
  • Context utilization
  • Response grounding

Troubleshooting​

Connection Failed​

  1. Verify Phoenix is running
  2. Check host URL is correct
  3. Ensure network connectivity

No Traces​

  1. Verify instrumentation is active
  2. Check Phoenix UI for traces
  3. Adjust time range settings

Next Steps​