← Browse all tools

Go Back
Report Abuse

Arize Phoenix

AI observability and evaluation platform

Tool overview

Phoenix is an AI observability platform from Arize designed for experimentation, evaluation and troubleshooting of LLM applications. It traces the runtime of an application with OpenTelemetry-based instrumentation.

Evaluations use LLMs to benchmark the responses and retrieval of an application. Versioned datasets of examples support experiments that track the effect of changes to prompts, models and retrieval. A playground lets users optimise prompts, compare models, adjust parameters and replay traced LLM calls, and prompt management versions and tests prompt changes.

Phoenix is free to use and self-host. Its source code is published under the Elastic License 2.0. Arize also offers Arize AX, a commercial platform for managed production workflows.

Screenshots & video

Arize Phoenix
Arize Phoenix

Community reviews

Login to Write Your Review

There are no reviews yet.

Source & attribution

Original contributor
Jean-Baptiste CROUIGNEAU
External references
https://github.com/Arize-ai/phoenix