CROP PDF Parser Service - Documentation

Complete documentation for the CROP PDF Parser Service system.

Overview

The CROP PDF Parser Service is an AI-powered parts search system for agricultural equipment manuals. It enables fast, accurate parts discovery using semantic search, RAG (Retrieval-Augmented Generation), and AI agents.

Key Features

• 🔍 Semantic Search: Natural language queries to find parts
• 📄 PDF Parsing: Automatic extraction of tables, schemas, and images
• 🤖 AI Agents: LangChain/LangGraph agents with LLaMA 3.1 8B
• 🖼️ Visual Search: CLIP-based image similarity search
• 📊 Vector Database: Weaviate for efficient RAG retrieval
• 💬 Chat Interface: React frontend with interactive PDF viewer
• 🔗 Store Integration: Direct links to parts store

Documentation

For Users

• API Query Format - How to query the AI agent
• Features Checklist - Complete feature list
• Architecture Guide - System architecture overview
• Frontend Integration - Frontend integration guide
• GCS PDF Links - PDF link generation

For Developers

See Developer Documentation for:

• Setup and deployment guides
• Technical implementation details
• Service-specific documentation
• Configuration and troubleshooting

Quick Start

API Usage

# Query the AI agent
curl -X POST "http://localhost:8000/api/ai/query" \
  -H "Content-Type: application/json" \
  -d '{
    "query": "Find part 10",
    "include_store_link": false
  }'

Frontend

The frontend is available at http://localhost:3000 (local) or deployed on Cloud Run (production).

System Architecture

The system consists of:

1. Monolithic Service - All core functionality in one service
2. Weaviate DB - Vector database for RAG
3. llama.cpp Server - LLM inference (CPU-optimized)
4. Frontend - React/TypeScript chat interface

For detailed architecture, see Architecture Guide.

API Endpoints

Main Endpoints

• POST /api/ai/query - Query AI agent
• POST /query - Alias for /api/ai/query
• GET /test-chat - Test chat endpoint
• GET /health - Health check

For complete API documentation, see API Query Format.

Support

For technical questions or issues, see Developer Documentation.