A full-stack web application that allows users to upload academic PDFs and ask natural language questions about the content using AI-powered retrieval augmented generation (RAG).

AI-Powered Research Assistant/
├── frontend/           # HTML, CSS, JavaScript interface
├── backend/           # Flask API server
├── ml/               # Machine learning modules
├── data/             # PDF storage and FAISS index
├── docs/             # Documentation
└── requirements.txt   # Python dependencies

• PDF Upload & Processing: Parse academic PDFs using PyMuPDF
• Semantic Search: Generate embeddings using OpenAI's text-embedding-ada-002
• RAG Implementation: Retrieve relevant chunks and generate answers with GPT-4
• Citation Generation: Automatic APA/MLA citation formatting
• Vector Storage: FAISS index for efficient similarity search
• Related Papers: Find similar papers using cosine similarity
• Azure OpenAI Support: Use Azure OpenAI endpoints and keys

• Python 3.8+
• OpenAI API key OR Azure OpenAI configuration
• pip (Python package manager)

1. Clone the repository

   git clone <repository-url>
   cd ai-research-assistant

2. Install dependencies

   pip install -r requirements.txt

3. Configure API Keys

   Option A: OpenAI Direct API

   export OPENAI_API_KEY="your-openai-api-key-here"

   Option B: Azure OpenAI

   export AZURE_OPENAI_ENDPOINT="https://your-resource.openai.azure.com/"
   export AZURE_OPENAI_API_KEY="your-azure-api-key-here"
   
   # Optional: Specify deployment names
   export AZURE_OPENAI_DEPLOYMENT_NAME="your-gpt-4-deployment"
   export AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME="your-embedding-deployment"

4. Run the application

   python backend/app.py

5. Access the application

   • Open your browser and navigate to http://localhost:5000

• Get your API key from OpenAI Platform
• Set environment variable: OPENAI_API_KEY

• Create an Azure OpenAI resource in Azure Portal
• Get your endpoint URL and API key
• Set environment variables:
  • AZURE_OPENAI_ENDPOINT
  • AZURE_OPENAI_API_KEY
  • AZURE_OPENAI_DEPLOYMENT_NAME (optional)
  • AZURE_OPENAI_EMBEDDING_DEPLOYMENT_NAME (optional)

• app.py: Main Flask application
• routes.py: API endpoints for upload, question-answering, and citations
• config.py: Configuration settings

• pdf_parser.py: PDF parsing and chunking
• embedding_generator.py: OpenAI/Azure embedding generation
• vector_search.py: FAISS index management and similarity search
• rag_engine.py: Retrieval Augmented Generation implementation
• citation_generator.py: APA/MLA citation formatting

• index.html: Main application interface
• style.css: Styling and responsive design
• script.js: Frontend JavaScript functionality

• uploads/: Stored PDF files
• embeddings/: FAISS index files
• metadata/: Document metadata storage

• POST /upload: Upload and process PDF files
• POST /ask: Ask questions about uploaded documents
• GET /documents: List uploaded documents
• GET /related/<doc_id>: Find related papers

1. Upload PDFs: Drag and drop or select academic PDF files
2. Ask Questions: Type natural language questions about the content
3. Get Answers: Receive AI-generated answers with citations
4. Explore Related Papers: Discover similar academic papers

• Store your API keys securely
• Implement proper file validation for PDF uploads
• Consider rate limiting for API endpoints
• Add authentication for production use

• Supports PDFs up to 50MB
• Processes documents in chunks of 1000 tokens
• Retrieves top-5 most relevant chunks for each question
• FAISS index enables fast similarity search

• GPT-4: ~$0.03 per 1K input tokens, ~$0.06 per 1K output tokens
• text-embedding-ada-002: ~$0.0001 per 1K tokens

• Pricing varies by region and model deployment
• Check Azure pricing calculator for your specific setup

1. Fork the repository
2. Create a feature branch
3. Make your changes
4. Submit a pull request

MIT License - see LICENSE file for details