agentic-rag-pydanticai

Agentic RAG with Pydantic AI

Consider giving this repo a ✨! Thanks!!!

Prerequisites

You need to have the following tools installed:

• uv
• Inferix or any OpenAI compatible API.

Environment Setup

1. Setup uv

# To setup uv
uv sync

2. Install Inferix (OpenAI compatible API)

Note: You can use any OpenAI compatible backend or simply use OpenAI itself.

• Feel free to use any OpenAI compatible API.
• Make sure you have Ollama installed.
• Make sure you have pulled a model. I recommend Qwen 2.5 32B.
• Use this guide to setup inferix to host a OpenAI compatible API capable of function calling.

3. Set the environment variables

• Checkout the example dotenv file at .env.example.
• Create a new .env using the example one as a template.
• Replace the variables as necessary.

Run the app

We've got a few scripts in this project:

1. uv run src/00_humble_call.py - Runs the ingestion script to create the embeddings and full text search index.
2. uv run src/01-basic-rag.py - Demonstrates a basic RAG workflow.
3. uv run src/02a-rag-as-tool.py - Demonstrates how you can agents can perform RAG by means of a tool call.
4. uv run src/02b-rag-as-system-prompt.py - Demonstrates how you can use Pydantic AI's system prompt hook to perform RAG.
5. uv run src/02c-rag-combined.py - Demonstrates how you can combine the two approaches to perform RAG.
6. uv run src/03-agentic-rag.py - Demonstrates how you can use Pydantic AI's agentic workflow to perform RAG.