MLflow LlamaIndex Workflow Example

This example demonstrates how to build and optimize a Retrieval-Augmented Generation (RAG) workflow using LlamaIndex integrated with MLflow. The example covers various retrieval strategies such as vector search, BM25, and web search, along with logging, model tracking, and performance evaluation in MLflow.

Set Up

This repository contains a complete workflow definition, a hands-on notebook, and a sample dataset for running experiments. To clone it to your working environment, use the following command:

shell

git clone https://github.com/mlflow/mlflow.git

After cloning the repository, set up the virtual environment by running:

cd mlflow/examples/llama_index/workflow
chmod +x install.sh
./install.sh

Once the installation is complete, start Jupyter Notebook within the Poetry environment using:

poetry run jupyter notebook