This repository accompanies the blog post Multi-agentic RAG with Hugging Face Code Agents. This code is meant for educational purposes only. Here is the code
Code agents work by executing Python code, make sure you run the code in an isolated environment and don't allow any unsafe import or function. Always supervise the system execution inspecting the logs, and interrupt the execution if the system is trying to execute code that may pose risks.
As all Large Language Model based systems, this is also prone to hallucinations. Always check the sources used to generate the answers and don't trust the generated answers blindly.
The default embedding model for similarity retrieval is all-mpnet-base-v2, check the corresponding model card.