weiwill88/Local_Pdf_Chat_RAG

Transparent Python RAG reference with FAISS + BM25 hybrid retrieval, reranking, Gradio UI, and FastAPI.

What it solves

This project provides a transparent, modular Python implementation of a Retrieval-Augmented Generation (RAG) pipeline. It is designed as an educational tool for developers to learn, inspect, and experiment with how documents are processed and retrieved to provide context for LLMs, rather than being a production-ready service.

How it works

The system processes documents through a structured pipeline: parsing, chunking, and creating embeddings. It then employs a hybrid retrieval strategy, combining FAISS for dense retrieval and BM25 for keyword-based retrieval. Results can be further refined using an optional reranking step (via CrossEncoder or model-based scoring) before the final context is built and passed to an LLM for answer generation.

Who it’s for

Developers and students who want a reference implementation to understand the inner workings of RAG pipelines and those who wish to experiment with different model backends.

Highlights

  • Hybrid Retrieval: Combines FAISS and BM25 to improve search accuracy.
  • Modular Design: Core components (loader, splitter, embeddings, retriever, reranker, generator) are split into replaceable modules.
  • Flexible Backends: Supports local Ollama, SiliconFlow, and OpenAI-compatible APIs.
  • Broad Document Support: Handles PDF, TXT, Markdown, DOCX, XLS/XLSX, and PPTX files.
  • Dual Interfaces: Includes both a Gradio web UI and a FastAPI REST API.

Related

  • Project
    Bessouat40/RAGLight

    A lightweight and modular Python library for implementing Retrieval-Augmented Generation (RAG), allowing users to easily integrate various LLMs and vector stores to chat with their documents.

  • Project
    ray-project/llm-applications

    A comprehensive guide and implementation for building, scaling, and productionizing RAG-based LLM applications using the Ray framework.

  • Project
    DataScienceUIBK/Rankify

    A comprehensive Python toolkit for retrieval, re-ranking, and RAG, providing a unified interface to experiment with and deploy state-of-the-art models.

  • Project
    NVIDIA-AI-Blueprints/rag

    A reference solution for building enterprise RAG pipelines using NVIDIA NIM microservices, supporting multimodal data extraction and agentic multi-hop retrieval.

  • Project
    infiniflow/ragflow

    An open-source RAG engine that combines deep document understanding and agentic capabilities to create a high-fidelity context layer for LLMs using complex unstructured data.