LLM Document Chat

Take-Home Project: LLM-Powered Data Extraction & Summarization

A Next.js service that lets users upload documents, automatically classifies document types (invoices, purchase orders, receipts, etc.), and extracts key structured data using OpenAI's Vision API. The extracted data is stored in a database and accessible via retrieval APIs.

Features

Core Requirements ✅

• Document Upload: File upload mechanism supporting PDFs, images, text files, and CSV
• Document Classification: Automatically classifies documents as invoices, purchase orders, receipts, contracts, reports, forms, letters, or other
• Structured Data Extraction: Extracts relevant fields based on document type:
  • Invoices: Invoice number, date, vendor, total, tax, bill-to information, line items
  • Purchase Orders: PO number, date, vendor, buyer, total, items, delivery date, terms
  • Receipts: Store, date, total, tax, payment method, items
  • Contracts: Parties, date, title, value, terms, duration
• Database Storage: All extracted data saved to SQLite database with Prisma ORM
• Retrieval APIs: Complete set of endpoints for document management and search

Additional Features

• Vector Search: Semantic search across document content using OpenAI vector stores
• Real-time Progress: SSE-based progress tracking for document processing
• Structured Logging: Comprehensive logging with Pino for debugging and monitoring
• Type Safety: Full TypeScript implementation with Zod validation
• Testing: Jest test suite with API route coverage

Getting Started

Prerequisites

• Node.js 18+ and npm
• OpenAI API key
• (Optional) Poppler tools for PDF processing (pdftocairo)
• (Optional) Redis/Upstash for caching

Setup

1. Clone the repository:

git clone https://github.com/wolyslager/llm-document-chat.git
cd llm-document-chat

2. Install dependencies:

npm install

3. Set up environment variables:

cp env.example .env.local
# Edit .env.local with your OpenAI API key and other settings

4. Set up the database:

npx prisma migrate dev --name init

5. Create a vector store:

node setup-vector-store.js

6. Run the development server:

npm run dev

Open http://localhost:3000 with your browser to see the application.

Project Structure

• src/app/api/ - Next.js API routes for document upload, search, and management
• src/lib/ - Core utilities (OpenAI, database, logging, validation)
• src/components/ - React components for the UI
• scripts/ - Utility scripts for cleanup and setup
• prisma/ - Database schema and migrations
• src/__tests__/ - Jest test suite

API Routes

Document Processing

• POST /api/upload - Upload documents for classification and data extraction

Retrieval APIs (Core Requirement)

• GET /api/documents - Fetch list of uploaded documents with metadata and extracted data summaries
• GET /api/documents/[id] - Fetch document with all relevant information including full extracted data
• POST /api/search - Basic search across documents using semantic vector search
• POST /api/search - Ask generic questions about documents via natural language queries

Vector Store Management

• GET /api/vector-stores - List available vector stores
• POST /api/vector-stores - Create new vector store
• DELETE /api/documents/[id] - Remove documents and clean up associated data

Scripts

• npm run dev - Start development server
• npm run build - Build for production
• npm run test - Run test suite
• npm run lint - Run ESLint
• npm run type-check - Run TypeScript compiler
• ./scripts/cleanup-all.js - Clean database and vector stores
• ./scripts/cleanup-database.js - Clean database only
• ./scripts/cleanup-vector-store.js - Clean vector store only

Implementation Approach

This solution demonstrates clean, well-structured code following best practices as requested:

1. Modular Architecture: Separation of concerns with dedicated modules for database, OpenAI integration, validation, and error handling
2. Type Safety: Full TypeScript implementation with Zod schemas for runtime validation
3. Error Handling: Unified error handling with structured JSON responses and proper HTTP status codes
4. Testing: Comprehensive Jest test suite covering API routes and core functionality
5. Logging: Structured logging with Pino for debugging and monitoring
6. Code Quality: ESLint configuration and consistent coding patterns

Technologies

• Framework: Next.js 14 with App Router
• Database: SQLite with Prisma ORM
• AI/ML: OpenAI GPT-4 Vision API, Vector Stores, Assistants API
• Validation: Zod schemas
• Logging: Pino structured logging
• Testing: Jest with API route testing
• Styling: Tailwind CSS
• TypeScript: Full type safety throughout

License

MIT