Phoenix by Arize
by Arize AI
Open-source ML observability tool for debugging LLM, CV, and NLP models
Visit Product
349 upvotes
374 views
About
Phoenix by Arize is an open-source machine learning observability platform that helps data scientists and ML engineers debug and improve models in production and during development. It provides deep visibility into model behavior through embedding analysis, drift detection, and performance monitoring — making it possible to understand why a model is underperforming and exactly what data is causing problems.
For LLM applications, Phoenix excels at tracing entire application flows — including retrieval steps, embedding operations, and model calls — surfacing root causes of quality issues. Its embedding visualizer allows teams to visually explore how data clusters in high-dimensional space, identify outliers, and understand model confidence across different input types.
Phoenix integrates with all major ML frameworks and LLM toolchains including LangChain, LlamaIndex, OpenAI, and Hugging Face. The open-source version runs locally in a Jupyter notebook or as a standalone server, while Arize's cloud platform provides production-grade monitoring at scale.
For LLM applications, Phoenix excels at tracing entire application flows — including retrieval steps, embedding operations, and model calls — surfacing root causes of quality issues. Its embedding visualizer allows teams to visually explore how data clusters in high-dimensional space, identify outliers, and understand model confidence across different input types.
Phoenix integrates with all major ML frameworks and LLM toolchains including LangChain, LlamaIndex, OpenAI, and Hugging Face. The open-source version runs locally in a Jupyter notebook or as a standalone server, while Arize's cloud platform provides production-grade monitoring at scale.
Product Features
- LLM application tracing and debugging
- Embedding visualization in 2D/3D space
- Data drift detection across production datasets
- Performance monitoring with custom metrics
- Root cause analysis for model degradation
- OpenTelemetry-compatible instrumentation
- LangChain and LlamaIndex native integrations
- Computer vision and NLP model support
- Interactive UI for data exploration
- Self-hosted open-source version
- Embedding visualization in 2D/3D space
- Data drift detection across production datasets
- Performance monitoring with custom metrics
- Root cause analysis for model degradation
- OpenTelemetry-compatible instrumentation
- LangChain and LlamaIndex native integrations
- Computer vision and NLP model support
- Interactive UI for data exploration
- Self-hosted open-source version
About the Publisher
Arize AI was founded in 2020 by Jason Lopatecki and Aparna Dhinakaran, both veterans of the machine learning industry. Headquartered in San Francisco, Arize has raised over $38 million in funding and serves ML teams at major enterprises across finance, technology, and healthcare. Phoenix, the company's open-source observability tool, was released to bring the power of production ML monitoring to teams that can't afford enterprise software, reflecting Arize's commitment to improving the entire ML ecosystem.