Phoenix Tracing

What it does

This skill enables AI agents to integrate with Arize Phoenix, a platform designed for observability and monitoring of machine learning systems. It allows agents to query performance metrics, analyze model drift, and visualize data pipelines directly within their operational context.

When to use it

• Investigating sudden drops in model accuracy or prediction confidence scores.
• Monitoring real-time data quality issues feeding into a training pipeline.
• Generating visual reports on feature distributions and label consistency over time.
• Debugging complex ML workflows where traditional logging is insufficient.

Key capabilities

• Connects AI agents to the Arize Phoenix observability platform.
• Retrieves detailed metrics regarding model performance and health.
• Facilitates deep dives into data lineage and pipeline execution history.
• Supports visualization of model behavior across different environments.

Example prompts

• "Check the current drift status for our customer churn prediction model in production."
• "Summarize the last 24 hours of data quality issues detected by Phoenix."
• "Visualize the feature importance distribution for the latest batch of training data."

Tips & gotchas

Ensure your AI agent has proper authentication credentials to access the Arize Phoenix instance before attempting queries. This skill is specific to the Phoenix ecosystem and will not function with other observability tools like Prometheus or Datadog without additional configuration.