phoenix.arize.com - Open-source LLM tracing and evaluation

phoenix.arize.com Open-source LLM tracing and evaluation

What is phoenix.arize.com?

Phoenix offers flexible, transparent, and free-from-lock-in AI observability. It utilizes OpenTelemetry (OTEL) for seamless setup and provides total visibility into application tracing. Leverage the interactive prompt playground and streamlined evaluations, complete with pre-built templates and human feedback options.

Phoenix supports dataset clustering and visualization, leveraging embeddings to uncover semantically similar data points and isolate performance issues. Proudly open source, it offers compatibility with numerous LLM tools, ensuring flexibility in the generative AI landscape.

Features

Application Tracing: Collect LLM app data with automatic or manual instrumentation for total visibility.
Interactive Prompt Playground: Experiment with prompts using pre-built templates or custom configurations.
Streamlined Evaluations: Use a fast and ergonomic eval library, incorporating pre-built templates, custom tasks, and human feedback.
Dataset Clustering & Visualization: Identify performance issues by analyzing semantically similar questions, document chunks, and responses.
OpenTelemetry Integration: Benefit from vendor-agnostic, framework-agnostic, and language-agnostic flexibility.