PDF Parser Service

FastAPI service for parsing PDF manuals and extracting text spans, tables, and schemas.

Features

• Extract text spans with coordinates using PyMuPDF
• Extract tables using Camelot
• Extract schema information (part numbers on diagrams)
• RESTful API with Swagger documentation

Quick Start

Local Development

# Install dependencies
pip install -r requirements.txt

# Run the service
uvicorn main:app --reload

The service will be available at http://localhost:8000

API Endpoints

• GET / - Health check
• GET /docs - Swagger documentation
• POST /parse - Parse PDF file and extract text spans
• POST /parse-table - Extract tables from PDF
• POST /prepare-rag - Prepare parsed PDF data for RAG system (Weaviate)

Example Usage

Parse PDF:

curl -X POST "http://localhost:8000/parse" \
  -H "accept: application/json" \
  -F "file=@example.pdf"

Extract Tables:

curl -X POST "http://localhost:8000/parse-table" \
  -H "accept: application/json" \
  -F "file=@example.pdf" \
  -F "page=0"

Prepare RAG Data:

curl -X POST "http://localhost:8000/prepare-rag" \
  -H "Content-Type: application/json" \
  -d '{
    "tables": [{"page": 6, "tables": [{"rows": [["Part", "Description"], ["10", "WASHER"]], "bbox": [10, 20, 90, 80]}]}],
    "schemas": [{"page": 6, "spans": [{"text": "10", "bbox": [11.76, 9.09, 29.41, 9.85]}]}],
    "source_pdf": "example.pdf"
  }'

The /prepare-rag endpoint converts parsed PDF data into RAG-ready format for Weaviate vector database ingestion. It processes tables and schemas from /parse and /parse-table endpoints and returns documents with content and metadata structure.

Deployment

Quick Deploy

cd pdf_parser_service
./deploy.sh

Manual Deploy

cd pdf_parser_service
gcloud builds submit --config cloudbuild.yaml .

Using Root Deploy Script

./deploy.sh pdf-parser

See deployment section in README.md for full deployment instructions.

Dependencies

• FastAPI - Web framework
• PyMuPDF (fitz) - PDF parsing
• Camelot - Table extraction
• Pandas - Data processing