A FastAPI-based backend service that processes and analyzes Informed Consent documents for human subjects research. The service uses advanced language models and embeddings to generate clear, understandable summaries of complex research consent documents.
- PDF document processing and analysis
- Hierarchical document parsing using LlamaIndex
- Concurrent processing of document sections
- OpenAI GPT-4o integration for intelligent summarization
- Specialized in bioethics and patient advocacy context
- RESTful API endpoints for document upload and processing
- Python 3.x
- OpenAI API key (environment variable:
OPENAI_API_KEY)
- Clone the repository:
git clone [repository-url]
cd irb-ki-backend- Install dependencies:
pip install -r requirements.txt- Set up environment variables:
export OPENAI_API_KEY='your-api-key'GET /: Health check endpoint- Response:
{"Hello": "World3"}
POST /uploadfile/: Upload and process PDF documents- Accepts: PDF file upload
- Returns: JSON object containing sectioned summaries of the document
The application uses a multi-layered approach to process documents:
- Document Ingestion: PDF documents are processed using custom PDF reader
- Text Processing: Documents are parsed into hierarchical nodes
- Embedding & Indexing: Text is embedded using OpenAI's text-embedding-3-large model
- Query Processing: Uses LlamaIndex for efficient document querying
- Summary Generation: Employs GPT-4o for generating clear, authoritative summaries
The project follows these key principles:
- Concurrent processing for improved performance
- Hierarchical document parsing for better context understanding
- Error handling and input validation
- CORS middleware for frontend integration
Key dependencies include:
- fastapi[standard] >= 0.113.0
- pydantic >= 2.7.0
- python-multipart
- pypdf
- llama_index
Start the FastAPI server:
uvicorn app.main:app --reloadThe server will start on http://localhost:8000