AI DocAssistant

A Python-based tool designed to extract key information from PDF documents using AI-powered models. The system parses documents, retrieves relevant sections, and extracts the data points in a structured format.

Project Structure

The core components of the project are:

• read: Parses any PDF and identifies different sections (headers, text, images, etc.).
• retrieve: Retrieves the relevant document chunks based on specified data points.
• extract: Extracts key data points from the relevant chunks, along with bounding box coordinates and references.
• api: Serves the product through an API, which takes PDFs as input and returns extracted data points.

doc-assistant/
│
├── api/                         # API routes and entry point
├── doc_assistant/
│   ├── pipelines/               # folder for different document pipelines
│   │   ├── epc.py               # Energy Performance Certificate pipeline
│   │   ├── invoice.py           # Invoice pipeline (can be added)
│   │   └── ...
│   ├── extract/                 # Extraction logic (e.g., extractor functions)
│   ├── read/                    # Parsing/reading logic
│   ├── retrieve/                # Logic to retrieve relevant chunks of text
│   ├── shared/                  # Shared entities and logic
│   └── config/                  # Configuration files for various document types
├── .env.sample                  # Template for environment variables
├── Makefile                     # Automation through make commands
├── README.md                    # Documentation
├── requirements.txt             # Python dependencies

Requirements

• Python 3.10
• Virtual environment (venv)
• Libraries specified in requirements.in

Installation

1. Clone the repository:

   git clone https://github.com/.git
   cd gide_rag

2. Set up a virtual environment: Ensure Python 3.10 is installed, and create a virtual environment using the Makefile:

   make create-env

3. Activate the virtual environment: On macOS/Linux:

   source venv/bin/activate

   On Windows:

   venv\Scripts\activate

Running Tests

To run backend tests using pytest, use:

make test

API Usage

1. Run the FastAPI backend server:

make run-backend

2. Test the API with curl:

curl -X 'POST' \
  'http://127.0.0.1:8000/extract/' \
  -H 'accept: application/json' \
  -H 'Content-Type: multipart/form-data' \
  -F 'document_type=epc' \
  -F 'language=en' \
  -F 'file=@doc_assistant/test_data/energy_performance_certificate_example.pdf'