Skip to content

README.md

Latest commit

 

History

History
133 lines (97 loc) · 3.79 KB

README.md

File metadata and controls

133 lines (97 loc) · 3.79 KB

PDF Analyzer with Local Ollama

A powerful tool to analyze PDF documents and answer questions using Retrieval-Augmented Generation (RAG) with your local Ollama installation. This tool extracts content only from the provided PDF and doesn't use any external knowledge sources.

Features

  • PDF Text Extraction: Uses pdfplumber for robust text extraction, including Arabic text
  • Semantic Search: Creates embeddings and uses FAISS for fast similarity search
  • Local AI: Uses your local Ollama installation for answering questions
  • Multi-language Support: Works with Arabic, English, and other languages
  • Two Interfaces: Both web UI (Streamlit) and command-line interface
  • Source Citation: Shows which parts of the PDF were used to generate answers

Installation

  1. Install Python dependencies:
pip install -r requirements.txt

  1. Install and setup Ollama:

    • Download Ollama from https://ollama.ai
    • Install and start the Ollama service
    • Pull a model (e.g., ollama pull llama3.2)

  2. Verify Ollama is running:

ollama list

Usage

Web Interface (Recommended)

  1. Start the Streamlit app:
streamlit run pdf_analyzer.py

  1. Open your browser to http://localhost:8501

  2. Upload your PDF and click "Process PDF"

  3. Ask questions about the PDF content

Command Line Interface

  1. Run the CLI:
python cli.py "path/to/your/pdf/file.pdf"
  1. Ask questions interactively

Advanced Options

python cli.py "document.pdf" --model llama3.2 --chunk-size 500 --overlap 100 --top-k 5

Configuration

Edit config.py to customize:

  • Ollama model name
  • Chunk size and overlap
  • Embedding model
  • Number of relevant chunks to retrieve

How It Works

  1. Text Extraction: Extracts text from PDF using pdfplumber
  2. Text Chunking: Splits text into overlapping chunks for better context
  3. Embedding Creation: Creates vector embeddings using SentenceTransformers
  4. Vector Storage: Stores embeddings in FAISS index for fast similarity search
  5. Question Processing:
    • Converts question to embedding
    • Finds most similar text chunks
    • Sends relevant context to Ollama
    • Returns AI-generated answer based only on PDF content

Supported Models

Any Ollama model can be used. Popular choices:

  • llama3.2 (recommended for general use)
  • mistral
  • codellama (for code-related documents)
  • qwen2.5 (good for multilingual content)

Make sure to pull the model first: ollama pull model-name

Troubleshooting

Ollama Connection Issues

  • Ensure Ollama is running: ollama serve
  • Check if model is available: ollama list
  • Verify the model name in configuration

PDF Processing Issues

  • Ensure the PDF contains extractable text (not just images)
  • Try with a different PDF to isolate the issue
  • Check file permissions

Memory Issues

  • Reduce chunk size in configuration
  • Use a smaller embedding model
  • Process smaller PDFs

Example Questions

For an Arabic legal document:

  • "ما هو موضوع هذا النظام؟" (What is the subject of this system?)
  • "ما هي المواد المتعلقة بالحكم؟" (What are the articles related to governance?)

For English documents:

  • "What is the main topic of this document?"
  • "Summarize the key points"
  • "What are the requirements mentioned?"

File Structure

pdf_analyzer/
├── pdf_analyzer.py      # Main Streamlit application
├── cli.py              # Command-line interface
├── config.py           # Configuration settings
├── requirements.txt    # Python dependencies
├── README.md          # This file
└── cache/             # Cached indexes (created automatically)

License

This project is open source and available under the MIT License.