Phasellm

PhaseLLM is a toolkit for building and testing products powered by large language models. It has two parts: an open source Python library, installed with pip, and a hosted platform for logging and evaluation. The library wraps several model providers behind one interface so teams can swap between GPT models, Claude, Hugging Face models and Google's Gemini and other Vertex AI models, and it supports Azure hosted OpenAI endpoints. It also includes a web search agent, a way to run conversations between two models, and utilities for prompt optimization and performance monitoring. The hosted platform lets teams log chats directly from their applications, then review and experiment with them without writing code. PhaseLLM Evals is described as a prompt engineering copilot: it runs batch tests of new models, new responses and revised prompts, and an experimental chained tables feature links columns of data to build multi step test workflows. The site also hosts tutorials with code, covering topics such as a news summarizer that cites sources and a smart RSS reader for new research papers. It is aimed at developers and product teams shipping LLM features who want a provider neutral layer plus a place to evaluate prompts. The site offers sign up and log in for the hosted platform but does not publish prices, so cost details for the platform are not listed.