Phoenix by Arize

Phoenix is a comprehensive tool for LLM observability that enables developers and data scientists to monitor, evaluate, and optimize their AI models. It captures detailed telemetry of LLM workflows, allowing teams to trace requests, visualize execution paths, detect failures, and debug model behavior effectively. The platform includes features such as an interactive prompt playground, pre-built and customizable evaluation templates, dataset clustering, and human feedback integration. Phoenix is compatible with popular frameworks like LangChain, LlamaIndex, and OpenAI, and is fully open-source and self-hostable, allowing flexibility and avoiding vendor lock-in. It improves model reliability, performance, and output quality in production environments.