Skip to content

CLAUDE.md

Latest commit

 

History

History
119 lines (88 loc) · 3.98 KB

CLAUDE.md

File metadata and controls

119 lines (88 loc) · 3.98 KB

CLAUDE.md

This file provides guidance to Claude Code (claude.ai/code) when working with code in this repository.

Project Overview

ChittyTrace is a comprehensive AI-powered financial forensics platform for analyzing documents, tracing fund flows, generating court-ready exhibits, and providing intelligent financial analysis capabilities. It integrates with Neon PostgreSQL databases and supports Cook County court filing requirements.

Key Commands

Development & Testing

# Activate virtual environment
source venv/bin/activate  # On Windows: venv\Scripts\activate

# Run the Streamlit GUI application
streamlit run app.py

# Start the API server for Claude/OpenAI extension
python claude_openai_extension.py

# Run tests (when implemented)
pytest tests/

Linting & Type Checking

# Run linting
ruff check .

# Run type checking
mypy flow_analyzer/

Architecture Overview

Core Components

  1. document_processor.py: Handles document scanning, text extraction, and caching

    • Supports PDF, Excel, CSV, text files
    • Implements intelligent caching system
    • Extracts metadata and content

  2. claude_integration.py: Claude API integration and analysis engine

    • Vector store for semantic search using Chroma
    • Complex command execution (fund tracing, timeline generation)
    • Form filling and exhibit generation

  3. recursive_scanner.py: Enhanced document discovery

    • Recursively scans directories for all document types
    • Integrates with Cloudflare Workers for email ingestion
    • Handles archives and communication files

  4. interactive_timeline.py: Timeline visualization system

    • Creates interactive Plotly timelines
    • Links timeline events to source documents
    • Extracts events using Claude AI

  5. package_generator.py: Cook County exhibit package generator

    • Formats exhibits per Cook County requirements
    • Generates cover sheets, affidavits, and indices
    • Creates complete filing packages

  6. database_handler.py: Neon PostgreSQL integration

    • Vector search capabilities
    • Cross-references with connected Neon database
    • Stores documents, events, and analysis results

  7. claude_openai_extension.py: API interface for AI assistants

    • FastAPI-based REST API
    • OpenAI function calling schema
    • Claude tool use integration

  8. app.py: Streamlit GUI application

    • Six main tabs: Query, Package Creator, Form Filler, Commands, Visualizations, Browser
    • Real-time document analysis
    • Interactive visualizations

Database Integration

The system uses Neon PostgreSQL with the following key tables:

  • documents: Stores all scanned documents with embeddings
  • timeline_events: Financial events extracted from documents
  • court_exhibits: Tracks exhibit preparation
  • analysis_queries: Query history and AI responses
  • exhibit_packages: Complete court filing packages

Important: The system can cross-reference with connected Neon databases for matching transactions, property records, and legal filings when a database is configured.

Cook County Requirements

When generating exhibits, the system automatically applies:

  • 8.5" x 11" page size with 1" margins
  • Times New Roman or Arial 12pt font
  • Double spacing
  • Sequential exhibit numbering
  • Exhibit stickers (bottom right)
  • Required authentication affidavits

Email Ingestion

The system integrates with Cloudflare Workers to ingest emails from nick@chitty.cc. Configure via:

  • CLOUDFLARE_WORKER_URL: Worker endpoint
  • CLOUDFLARE_WORKER_TOKEN: Authentication token

Environment Variables

Required in .env:

  • ANTHROPIC_API_KEY: Claude API access
  • DATABASE_URL: Neon database connection (optional)
  • CLOUDFLARE_WORKER_URL: Email ingestion endpoint
  • API_TOKENS: Bearer tokens for API authentication

ChittyCorp Integration

This system integrates with ChittyCorp's enterprise AI infrastructure: