Can I use this skill to find the slowest part of my AI chain?

Yes, the skill can list recent traces sorted by duration and drill down into individual span latencies to pinpoint where delays are occurring in your application.

How do I view specific LLM trace details?

You can use the 'px trace ' command to fetch a detailed tree view or raw JSON data for a specific execution, including input/output values and metadata.

What is the Phoenix CLI skill for Claude Code?

It is a specialized capability that allows Claude to use the Arize Phoenix command-line tool to debug, trace, and evaluate the performance of LLM applications directly.

Does Phoenix CLI work with all LLM providers?

Yes, it works with any application instrumented with Phoenix or OpenInference-compatible libraries, regardless of whether you use OpenAI, Anthropic, or local models.

How do I filter for errors in my LLM application?

The skill allows you to fetch traces and filter them by status codes or use tools like jq to extract spans containing exception messages or non-OK statuses.

Phoenix AI Observability CLI

Name: Phoenix AI Observability CLI
Author: Arize-ai

byArize-ai

•

8,417

分析与监控

Debugs and optimizes LLM applications by analyzing traces, experiments, and evaluation datasets through the Phoenix CLI.

关于

The Phoenix CLI skill empowers Claude to interact directly with Arize Phoenix, an industry-leading observability platform for AI. It enables developers to troubleshoot failing LLM chains, identify performance bottlenecks such as high latency or excessive token usage, and review detailed experiment results without leaving the terminal. By utilizing OpenInference semantic attributes, the skill provides granular visibility into span data—including model inputs, outputs, and system errors—making it an essential tool for building, testing, and maintaining reliable AI applications.

主要功能

Performance bottleneck identification using latency and token metrics
8,417 GitHub stars
Flexible output formatting including pretty-print and raw JSON for automation
Support for OpenInference semantic attributes for standardized observability
Deep trace fetching and span analysis for complex LLM workflows
Experiment result review and dataset exploration for AI benchmarks

使用场景

Monitoring and optimizing token consumption and model costs across providers
Debugging failing LLM spans or chain errors in complex RAG pipelines
Analyzing experiment data to validate model improvements and prevent regressions

关于

主要功能

Performance bottleneck identification using latency and token metrics
8,417 GitHub stars
Flexible output formatting including pretty-print and raw JSON for automation
Support for OpenInference semantic attributes for standardized observability
Deep trace fetching and span analysis for complex LLM workflows
Experiment result review and dataset exploration for AI benchmarks

使用场景

Monitoring and optimizing token consumption and model costs across providers
Debugging failing LLM spans or chain errors in complex RAG pipelines
Analyzing experiment data to validate model improvements and prevent regressions