LLM Analysis and Evaluation of LangChain and OpenAI RAG using Arize-Phoenix

Author(s): Steve George

Originally published on Towards AI.

This member-only story is on us. Upgrade to access all of Medium.

LLM workflows based on retrieval-augmented generation (RAG) are currently being used at scale in production environments. While token usage and distribution are tracked at the API level, it is highly recommended that these models be analyzed and evaluated. Obtaining ground truth values for evaluation may not always be feasible due to the enormous and exhausting nature of the task. Therefore, in real-world scenarios, another LLM model is often used to evaluate the workflow. This approach is called “LLM-as-a-judge” and is one of the most widely used methods.

Arize is one of the leading monitoring tools from both MLOps and LLMOps perspectives. Phoenix, an open-source observability library developed by Arize AI, is designed for experimentation, evaluation, and model troubleshooting.

In this article, we will leverage the Phoenix library to analyze and evaluate a LangChain RAG workflow.

Let’s start with installing all the required libraries.

!pip install langchain langchain-community langchain-openai openai "arize-phoenix[evals]" nest-asyncio chromadb

For content, we can create a dummy Dataframe to simulate the data. However, you’re encouraged to use the sample dataset provided by Arize or explore the extensive datasets available on Hugging Face, which offer more comprehensive real-world data for better analysis and… Read the full blog for free on Medium.

Join thousands of data leaders on the AI newsletter. Join over 80,000 subscribers and keep up to date with the latest developments in AI. From research to projects and ideas. If you are building an AI startup, an AI-related product, or a service, we invite you to consider becoming a sponsor.

Published via Towards AI