RAG-ecosystem

A complete Retrieval-Augmented Generation (RAG) system, plus an OCR-to-Markdown & Image Q&A module, built with Streamlit and Haystack.

Table of Contents

• Features
• Getting Started
  • Prerequisites
  • Installation
  • Configuration
• Usage
  • 1. Agentic RAG Chat & BM25 Search
  • 2. OCR to Markdown & Image Q&A
• Project Structure
• Contributing
• License

Features

1. Agentic RAG Chat

   • Hybrid retrieval (OpenAI embeddings + BM25)
   • Contextual query routing & summarization
   • Stateful chat sessions saved to SQLite

2. BM25-Only Document Search

   • Quick keyword-driven document lookup

3. OCR to Markdown Converter

   • Uses Together AI vision models to extract full-page content as Markdown

4. Image Question-Answering

   • Ask natural-language questions about any uploaded or URL’d image

Getting Started

Prerequisites

• Python 3.8+
• An OpenAI API key
• A Together AI API key (for OCR & Image Q&A)
• Git & your favorite terminal/shell

Installation

1. Clone the repo

   git clone https://github.com/Esmail-ibraheem/RAG-ecosystem.git
   cd RAG-ecosystem

2. Create & activate a virtual environment

   python3 -m venv .venv
   source .venv/bin/activate    # macOS/Linux
   .venv\Scripts\activate       # Windows

3. Install dependencies

   pip install streamlit
   pip install "haystack[all]"   # Core RAG components
   pip install openai sqlalchemy pandas python-docx pillow requests python-dotenv

Configuration

Create a .env file in the project root with your API keys:

OPENAI_API_KEY=sk-…
TOGETHER_API_KEY=sk-…

Or set them in your shell:

export OPENAI_API_KEY=sk-…
export TOGETHER_API_KEY=sk-…

Usage

1. Agentic RAG Chat & BM25 Search

Launch the main RAG app:

streamlit run RAG.py

• Sidebar

  • Enter your OpenAI API key.
  • Pick a GPT model (e.g. gpt-3.5-turbo or gpt-4-turbo).
  • Upload documents (.pdf, .docx, .txt, .csv, .xlsx) for RAG or BM25.
  • Start or load chat sessions.

• Main panel

  • For “RAG Chat”: chat with the system—behind the scenes it chooses between summary, context-driven answer, or simple reply.
  • For “BM25 Search”: run keyword searches and preview top‐k results.

All chat history is stored in chat_history.db (SQLite) for later reuse.

2. OCR to Markdown & Image Q&A

Run the OCR & QA utility:

streamlit run ocr_processor.py

• Convert to Markdown

  • Upload or URL-point to an image.
  • Click Convert to Markdown to get full‐page Markdown.

• Ask About the Image

  • Enter a natural-language question.
  • Click Get Answer to see the model’s response.

Project Structure

.
├── RAG.py                 # Agentic RAG & BM25 Search Streamlit app
├── ocr_processor.py       # OCR → Markdown & Image Q&A Streamlit app
├── utils/
│   └── custom_converters.py  # Docx/.xlsx → Haystack Document converters
├── chat_history.db        # SQLite DB (auto-generated)
├── LICENSE                # MIT License
└── .gitignore

Contributing

1. Fork the repo
2. Create a feature branch (git checkout -b feat/my-feature)
3. Commit your changes (git commit -m "Add …")
4. Push to your branch (git push)
5. Open a Pull Request!

Please follow the existing code style and include tests/examples where applicable.

License

This project is licensed under the MIT License. See the LICENSE file for details.