PDF RAG Application with FastMCP, Ollama, and LangChain

A simple application for querying PDF documents using Retrieval-Augmented Generation (RAG) with local LLMs via Ollama, with a fallback to the Financial Times API for financial data.

How it Works

1. Load PDFs: The app reads PDF files from a data/ folder.
2. Chunking: Text from PDFs is split into smaller, manageable chunks.
3. Embedding: Ollama converts these text chunks into numerical representations (embeddings).
4. Vector Store: Embeddings are stored in a FAISS vector store for semantic search.
5. Querying: When a question is asked (e.g., the default question), the app:
   - Converts the question to an embedding.
   - Finds the most relevant text chunks from the PDF.
   - Sends these chunks and the question to the Ollama LLM.
   - Returns an answer based on the PDF context.
   - If no answer is found in the PDF, it queries the Financial Times API for relevant data.
6. Output: The answer and its sources (PDF pages or FT API) are printed to the console.

Prerequisites

• Python 3.11: Download Python 3.11
• Poetry: For Python package management. Installation instructions
• Ollama: For running local LLMs.
• Download from ollama.com/download
• Pull the model: ollama pull llama3
• PDF File: A PDF file (e.g., Barclays-PLC-Annual-Report-2020.pdf) in the data/ folder.

Setup Instructions

1. Clone/Download the Project:
   - Ensure all files (main.py, pyproject.toml, .env.example) are in the project root.

2. Install Dependencies:
   - Open a terminal in the project root and run:
   bash poetry env use python3.11 poetry install