Logging and Evals for LLMs

Log, review, and evaluate your LLMs with our hosted platform

Learn more »

Open Source Product Tools

Use the phasellm Python library to build LLM-powered products

View GitHub repo »

PhaseLLM helps you build robust LLM-driven software products

Swap LLMs: GPT-4, Claude, Hugging Face, and more
Optimize prompts and monitor performance
Log chats directly to our hosted platform; review and experiment without coding

Installation Instructions

pip install phasellm

Read the Docs » GitHub »

Our Research

Gemini Support in PhaseLLM v0.0.21

PhaseLLM now supports Gemini and other Vertex AI models from Google. This post provides advice around how to set this up and use the API. Read More

Chained Tables (for LLM workflows)

We're releasing an experimental feature called "chained tables" for PhaseLLM Evals. This allows you to chain columns of data to test your prompts, reference variables, and build workflows. Read More

What does the perfect LLM product dev tool look like?

A list of features we think a perfect (or, very good!) LLM product development tool should have. Read More

Introducing PhaseLLM Evals

Evals is your prompt engineering copilot. PhaseLLM Evals helps you quickly scale your LLM experiments and tests. Run batch processes to test new models, new responses, and iterate on your prompts. Read More

Using LLMs to Build Smart RSS Readers (tutorial)

Build smart RSS readers with PhaseLLM. In this tutorial, we show you how to build a smart RSS reader to alert you about new ArXiv papers that you might be interested in. Code included! Read More

Using Microsoft Azure's OpenAI Endpoints

Leverage your organization's existing Microsoft Azure environment to call OpenAI endpoints. In this blogpost, we show you how to use PhaseLLM's new Azure support. Read More

Supercharge LLMs with Search (tutorial)

Use the PhaseLLM web search agent and an LLM to supercharge your llm chat experiences. In this tutorial, we enable LLMs to search the web to provide better answers to your questions. Read More

Structuring Data via LLM Web Crawling Agents

Use PhaseLLM webs search agents and LLMs to craw websites and build structured data sets. In this tutorial, we build a list of 200+ restaurants in Tokyo, along with descriptions and popularity counts. Code included! Read More

Roleplaying with LLMs (tutorial)

PhaseLLM makes it incredibly easy to set up a conversation between two LLMs. This is useful for roleplaying, training, and evaluation. This shows how to make your own roleplays and iterate on them. Code included! Read More

Newsbot: how to summarize news and cite sources

The Newsbot demo project summarizes news articles and cites sources, just like popular LLM-powered search engines. Learn how we evaluate and optimize prompts to enable this. Read More

How ResearchLLM works: architecture and data flows

This post provides an overview of how ResearchLLM is architected and how it works. Walks through the use case, prompt chaining, and how we manage models and errors. Read More

ResearchLLM: automated data analysis and interpretation

We're releasing our first demo product using PhaseLLM. ResearchLLM allows you to analyze data sets using natural language. It even interprets the results of its statistical code for you. Read More

Welcoming Dolly 2.0 with a Wrapper and Tests

DataBricks recently announced Dolly 2.0 and we've integrated it into PhaseLLM. Run Dolly 2.0 locally and within your existing PhaseLLM workflows. Read More

Contact Us

Want to collaborate? Have questions? Reach out to us at hello (at) phaseai (dot) com