diff --git a/.DS_Store b/.DS_Store new file mode 100644 index 0000000..bfe7b37 Binary files /dev/null and b/.DS_Store differ diff --git a/Dockerfile.allinone b/Dockerfile.allinone index f12dd1a..6cb376b 100644 --- a/Dockerfile.allinone +++ b/Dockerfile.allinone @@ -255,14 +255,22 @@ ENV CELERY_RESULT_BACKEND=redis://localhost:6379/0 ENV PYTHONPATH=/app/backend ENV NEXT_FRONTEND_URL=http://localhost:3000 ENV AUTH_TYPE=LOCAL -ENV ETL_SERVICE=DOCLING +ENV ETL_SERVICE=UNSTRUCTURED ENV EMBEDDING_MODEL=sentence-transformers/all-MiniLM-L6-v2 +# Asyncio loop configuration (required for unstructured/uvicorn compatibility) +ENV UVICORN_LOOP=asyncio +ENV UNSTRUCTURED_HAS_PATCHED_LOOP=1 + +# Disable LangSmith tracing by default (avoids SSL errors with corporate proxies) +ENV LANGCHAIN_TRACING_V2=false +ENV LANGSMITH_TRACING=false + # Frontend configuration (can be overridden at runtime) # These are injected into the Next.js build at container startup ENV NEXT_PUBLIC_FASTAPI_BACKEND_URL=http://localhost:8000 ENV NEXT_PUBLIC_FASTAPI_BACKEND_AUTH_TYPE=LOCAL -ENV NEXT_PUBLIC_ETL_SERVICE=DOCLING +ENV NEXT_PUBLIC_ETL_SERVICE=UNSTRUCTURED # Electric SQL configuration (ELECTRIC_DATABASE_URL is built dynamically by entrypoint from these values) ENV ELECTRIC_DB_USER=electric diff --git a/README.md b/README.md index 6e69f8c..5d8fb2e 100644 --- a/README.md +++ b/README.md @@ -2,9 +2,11 @@
-**Your AI-Powered Personal Finance Assistant** +FinanceGPT Logo -An intelligent financial management platform that helps you track spending, optimize rewards, analyze investments, and make smarter money decisions using AI. +### Your AI-Powered Personal CPA + +**Connect bank accounts • Upload tax forms • Get instant insights** [![License](https://img.shields.io/badge/License-Apache%202.0-blue.svg)](LICENSE) [![TypeScript](https://img.shields.io/badge/TypeScript-007ACC?logo=typescript&logoColor=white)](https://www.typescriptlang.org/) @@ -12,354 +14,179 @@ An intelligent financial management platform that helps you track spending, opti [![FastAPI](https://img.shields.io/badge/FastAPI-009688?logo=fastapi&logoColor=white)](https://fastapi.tiangolo.com/) [![Python](https://img.shields.io/badge/Python-3776AB?logo=python&logoColor=white)](https://www.python.org/) -[Features](#features) • [Quick Start](#quick-start) • [Architecture](#architecture) • [Documentation](#documentation) • [Contributing](#contributing) +[Quick Start](#-quick-start) • [Features](#-features) • [Example Prompts](#-example-prompts) • [Privacy](#-privacy-first-design) • [Contributing](#-contributing)
--- -## 🌟 Features +## 🚀 Quick Start -### 🤖 AI-Powered Financial Advisor -- **Smart Transaction Search**: Search your financial history using natural language queries - - "How much did I spend on restaurants last year?" - - "Show me all charges from United Airlines" - - "Find my recurring subscriptions" -- **Personalized Recommendations**: Get AI-driven suggestions for saving money and optimizing rewards -- **Natural Language Queries**: Ask questions about your finances in plain English -- **Predictive Analytics**: Forecast future spending and identify potential savings +Choose your preferred setup method: -### 💳 Smart Credit Card Optimization -- **Rewards Analysis**: Analyze your spending patterns to find the best credit cards - - "Which credit card should I use for restaurant purchases?" - - "Am I using the right credit card for my spending?" -- **Category-Based Optimization**: Get recommendations based on merchants and categories -- **Multi-Card Strategy**: Optimize rewards across multiple credit cards +### Option 1: All-in-One Docker (Easiest) -### 📈 Investment Portfolio Management -- **Real-Time Performance Tracking**: Monitor your investment returns with live market data - - "How are my stocks performing today?" - - "What's my portfolio return over the last year?" - - "Show my month-over-month performance" -- **Yahoo Finance Integration**: Fetches real-time and historical stock prices -- **Cost Basis Tracking**: Calculate unrealized gains/losses across all holdings -- **Time-Based Analysis**: Week-over-week, month-over-month, quarterly, and yearly performance - -### 🎯 Portfolio Allocation & Rebalancing -- **Asset Allocation Analysis**: Understand your portfolio composition (stocks/bonds/cash) - - "Is my portfolio allocation correct?" - - "How should I rebalance according to Bogleheads philosophy?" -- **Geographic Diversification**: Track US vs international exposure -- **Investment Philosophy Comparison**: Compare against established strategies - - Bogleheads Conservative (40/50/10) - - Bogleheads Moderate (60/35/5) - - Bogleheads Aggressive (90/10/0) - - Three-Fund Portfolio -- **Specific Rebalancing Recommendations**: Get dollar amounts for buying/selling -- **Alignment Score**: See how well your portfolio matches your target allocation (0-100) - -### 💰 Tax Optimization -- **Tax Loss Harvesting**: Identify opportunities to reduce your tax liability - - "Can I harvest any tax losses?" - - "What positions should I sell for tax savings?" -- **Loss Identification**: Finds holdings with unrealized losses -- **Tax Savings Calculator**: Estimates tax benefits based on your tax bracket -- **Replacement Suggestions**: Recommends similar securities to avoid wash sales -- **Wash Sale Warnings**: Alerts about IRS rules and compliance - -### 💳 Transaction & Spending Analysis -- **Multi-Account Aggregation**: Connect bank accounts, credit cards, and investment platforms via Plaid -- **Real-Time Tracking**: Monitor balances, transactions, and net worth in real-time -- **Subscription Detection**: Identify and track recurring payments automatically - - "Check if I have any recurring subscriptions" - - "What am I paying monthly?" -- **Category-Based Search**: Find transactions by category (restaurants, travel, etc.) -- **Merchant Search**: Search by merchant name with fuzzy matching - -### 📊 Analytics & Reporting -- **Interactive Dashboards**: Visualize spending trends, income, and investments -- **Custom Reports**: Generate detailed financial reports and summaries -- **Budget Management**: Set and monitor budgets with smart alerts -- **Historical Comparisons**: Compare spending and performance across time periods - -### 🔒 Security & Privacy -- **Bank-Level Encryption**: 256-bit SSL encryption for all data -- **Secure Authentication**: OAuth 2.0 and Google Sign-In support -- **Data Privacy**: Your financial data stays private and secure - -### 🔗 Integrations -- **100+ Financial Institutions**: Banks, credit cards, investment platforms, crypto exchanges -- **Plaid Integration**: Secure connection to financial accounts -- **Yahoo Finance**: Real-time stock prices and historical market data -- **Real-Time Sync**: Automatic transaction updates -- **Manual Uploads**: Support for CSV files (bank statements, Fidelity positions, etc.) -- **Export Options**: Download your data anytime +```bash +# Clone and run with a single container +git clone https://github.com/yourusername/FinanceGPT.git +cd FinanceGPT + +# Copy environment file and add your API keys +cp .env.example .env +# Edit .env with your OPENAI_API_KEY, PLAID_CLIENT_ID, PLAID_SECRET + +# Start FinanceGPT +docker compose -f docker-compose.quickstart.yml up -d +``` + +🎉 **Open http://localhost:3000** — You're done! --- -## 💬 Example Prompts +### Option 2: Local Development (macOS) -### Transaction Search & Analysis -``` -"How much did I spend on restaurants last year?" -"Show me all United Airlines charges" -"Find transactions over $100 in the last month" -"What did I spend on groceries this week?" -"Show my Amazon purchases" +```bash +# Clone the repository +git clone https://github.com/yourusername/FinanceGPT.git +cd FinanceGPT + +# Start infrastructure (PostgreSQL, Redis) +docker compose up -d db redis electric + +# Run the dev script (opens 3 terminal tabs automatically) +chmod +x dev.sh +./dev.sh ``` -### Credit Card Optimization +This starts: +- 🔧 **Backend API** on http://localhost:8000 +- 🔄 **Celery Worker** for background tasks +- 🌐 **Frontend** on http://localhost:3000 + +--- + +### Option 3: Full Docker Stack + +```bash +# Clone the repository +git clone https://github.com/yourusername/FinanceGPT.git +cd FinanceGPT + +# Configure environment +cp financegpt_backend/.env.example financegpt_backend/.env +# Edit .env with your API keys + +# Build and run all services +docker compose up -d --build + +# View logs +docker compose logs -f ``` -"Which credit card should I use for restaurants?" -"Am I using the right credit card for gas purchases?" -"What's the best card for my travel spending?" -"Optimize my credit card usage" + +--- + +## 🔐 Privacy-First Design + +Your financial data is sensitive. FinanceGPT is built with privacy as a core principle: + +| Feature | How It Protects You | +|---------|---------------------| +| **🔒 PII Masking** | SSN and EIN are **masked before any LLM call** (`123-45-6789` → `XXX-XX-XXXX`). Your tax forms never expose sensitive IDs. | +| **🏠 Self-Hostable** | Run entirely on your own hardware. Your data never leaves your machine. | +| **🤖 BYO Model** | Use your own LLM (OpenAI, Anthropic, or **local Ollama**). No vendor lock-in. | +| **🔐 Local Processing** | Sensitive field extraction (SSN, EIN) happens locally—not via cloud APIs. | +| **🗄️ Your Database** | All data stored in your PostgreSQL instance. Export or delete anytime. | +| **🚫 No Telemetry** | Zero tracking, zero analytics, zero data collection. | + +```python +# Example: How we handle your W2 +raw_text = "SSN: 123-45-6789, Wages: $183,000" +masked_text = mask_pii_in_text(raw_text) +# → "SSN: XXX-XX-XXXX, Wages: $183,000" +# Only masked_text is sent to the LLM ``` -### Investment Performance +--- + +## 💬 Example Prompts + +Just ask questions in plain English. FinanceGPT understands context. + +### 💰 Income & Tax Questions ``` -"How are my stocks performing today?" -"What's my portfolio return over the last year?" -"Show my month-over-month investment performance" -"How much have my investments grown this quarter?" -"What's my total portfolio value?" +"How much did I earn in 2024?" +"What was my total federal tax withheld?" +"Will I get a tax refund this year?" +"Show me my W2 summary" +"What state taxes did I pay?" ``` -### Portfolio Allocation & Rebalancing +### 📊 Spending Analysis ``` -"Is my portfolio allocation correct?" -"How should I rebalance according to Bogleheads philosophy?" -"What's my exposure to international stocks?" -"Am I too heavily invested in US stocks?" -"Should I buy more bonds or stocks?" -"Compare my portfolio to a three-fund strategy" +"How much did I spend on restaurants last month?" +"What are my recurring subscriptions?" +"Find all Amazon purchases over $100" +"What's my biggest expense category?" +"Show spending trends for the last 3 months" ``` -### Tax Loss Harvesting +### 💳 Credit Card Optimization ``` -"Can I harvest any tax losses?" -"What stocks should I sell for tax losses?" -"How much can I save in taxes by tax loss harvesting?" -"Are there any positions with unrealized losses?" -"Show me tax optimization opportunities" +"Which card should I use for groceries?" +"Am I using the right credit card for travel?" +"How much rewards am I missing out on?" +"Optimize my credit card usage" ``` -### Subscriptions & Recurring Payments +### 📈 Investment Portfolio ``` -"Check if I have any recurring subscriptions" -"What subscriptions am I paying for?" -"Find all my monthly recurring charges" -"Which services am I subscribed to?" +"How are my stocks performing today?" +"What's my portfolio return this year?" +"Is my allocation correct for my age?" +"Should I rebalance according to Bogleheads?" +"Can I harvest any tax losses?" ``` -### Financial Planning +### 🏦 Account Overview ``` "What's my net worth?" -"How much am I saving each month?" -"Show my spending trends over the last 3 months" -"What's my biggest expense category?" +"Show all my account balances" +"How much do I have in savings?" +"What's my monthly cash flow?" ``` --- -## �️ AI Tools & Capabilities - -FinanceGPT uses specialized AI tools to analyze your financial data and provide actionable insights: - -### 1. **Transaction Search** (`search_transactions`) -- Searches through all your financial transactions using keywords and categories -- Supports both Plaid-connected accounts and manual CSV uploads -- Fuzzy merchant name matching for accurate results -- Date range filtering and category-based filtering -- Returns transaction summaries with totals and breakdowns - -### 2. **Credit Card Optimizer** (`optimize_credit_card_usage`) -- Analyzes spending patterns to recommend optimal credit cards -- Matches merchant categories to card rewards programs -- Compares rewards rates across multiple cards -- Provides specific recommendations per spending category -- Supports both manual uploads and Plaid data - -### 3. **Portfolio Performance** (`calculate_portfolio_performance`) -- Fetches real-time stock prices from Yahoo Finance -- Calculates returns over custom time periods (day, week, month, quarter, year) -- Compares current prices to historical prices for accurate performance -- Shows individual holding performance and total portfolio returns -- Supports both snapshot comparisons and live price lookups - -### 4. **Portfolio Allocation Analyzer** (`analyze_portfolio_allocation`) -- Analyzes asset class distribution (stocks/bonds/cash) -- Calculates geographic exposure (US vs international) -- Compares portfolio to investment philosophies (Bogleheads, Three-Fund) -- Provides specific rebalancing recommendations with dollar amounts -- Gives alignment score (0-100) showing how close you are to target - -### 5. **Tax Loss Harvesting** (`analyze_tax_loss_harvesting`) -- Identifies positions with unrealized losses -- Calculates potential tax savings based on your tax bracket -- Suggests replacement securities to avoid wash sales -- Provides sell/buy recommendations for tax optimization -- Warns about wash sale rules and compliance - -### 6. **Subscription Finder** (`find_subscriptions`) -- Automatically detects recurring charges in your transaction history -- Identifies monthly, quarterly, and annual subscriptions -- Calculates total subscription costs -- Helps you find forgotten or unused subscriptions - -### 7. **Knowledge Base Search** (`search_knowledge_base`) -- Searches across all your uploaded financial documents -- Vector-based semantic search for accurate results -- Supports PDFs, CSVs, bank statements, and investment reports -- Context-aware retrieval for answering complex questions - ---- - -## �🚀 Quick Start - -### Option 1: Quick Start (Recommended for Testing) - -The fastest way to try FinanceGPT: - -```bash -# 1. Clone the repository -git clone https://github.com/yourusername/FinanceGPT.git -cd FinanceGPT - -# 2. Run the quick start script -chmod +x start-financegpt.sh -./start-financegpt.sh -``` - -This uses pre-built Docker images and requires minimal configuration. Access the app at **http://localhost:3000** - -### Option 2: Development Setup - -For development with full customization: - -#### Prerequisites - -- **Docker** and **Docker Compose** (required) -- **Node.js** 18+ and **pnpm** (for frontend development) -- **Python** 3.11+ (for backend development) - -#### Required Environment Variables - -Before starting, you **must** configure these essential variables: - -1. **Create backend environment file:** - ```bash - cp financegpt_backend/.env.example financegpt_backend/.env - ``` - -2. **Edit `financegpt_backend/.env` and set these required variables:** - - ```env - # === REQUIRED: LLM API Key === - # You MUST configure at least one LLM provider for FinanceGPT to work - # Get your API key from the respective provider: - - # OpenAI (recommended for best results) - OPENAI_API_KEY=sk-... - - # OR Anthropic Claude - ANTHROPIC_API_KEY=sk-ant-... - - # OR Google Gemini (free tier available) - GOOGLE_API_KEY=AIza... - - # === REQUIRED: Plaid (for bank account connections) === - # Sign up at https://dashboard.plaid.com/team/keys - PLAID_CLIENT_ID=your_plaid_client_id - PLAID_SECRET=your_plaid_secret - PLAID_ENV=sandbox # Use 'sandbox' for testing - - # === REQUIRED: Security === - SECRET_KEY=your-random-secret-key-change-this-in-production - - # === Optional: Document Processing === - # Only needed if you want to parse PDFs/documents - # Get free API key from https://unstructured.io/ - # UNSTRUCTURED_API_KEY=your_key_here - ``` - -3. **Create frontend environment file:** - ```bash - cp financegpt_web/.env.example financegpt_web/.env.local - ``` - - Edit `financegpt_web/.env.local`: - ```env - # Backend API URL - NEXT_PUBLIC_FASTAPI_BACKEND_URL=http://localhost:8000 - - # Auth type (LOCAL or GOOGLE) - NEXT_PUBLIC_FASTAPI_BACKEND_AUTH_TYPE=LOCAL - - # Document parsing service - NEXT_PUBLIC_ETL_SERVICE=DOCLING - ``` - -#### Installation Steps - -1. **Start infrastructure services (PostgreSQL, Redis, etc.)** - ```bash - docker-compose up -d db redis electric pgadmin - ``` - -2. **Install and run backend** - ```bash - cd financegpt_backend - - # Create virtual environment - python -m venv .venv - source .venv/bin/activate # On Windows: .venv\Scripts\activate - - # Install dependencies - pip install -e . - - # Run database migrations - alembic upgrade head - - # Start the backend server - uvicorn main:app --reload --host 0.0.0.0 --port 8000 - ``` - -3. **In a new terminal, install and run frontend** - ```bash - cd financegpt_web - - # Install dependencies - pnpm install - - # Start development server - pnpm dev - ``` - -4. **Access the application** - - 🌐 **Frontend**: http://localhost:3000 - - 🔧 **Backend API**: http://localhost:8000 - - 📚 **API Documentation**: http://localhost:8000/docs - - 🗄️ **pgAdmin** (Database UI): http://localhost:5050 - - Email: `admin@financegpt.com` - - Password: `financegpt` - -### Option 3: Full Docker Deployment - -Build and run everything with Docker: - -```bash -# Build and start all services -docker-compose up -d +## 🌟 Features -# View logs -docker-compose logs -f +### 🤖 AI-Powered Financial Advisor +- **Natural Language Queries**: Ask questions about your finances in plain English +- **Smart Transaction Search**: "How much did I spend on restaurants last year?" +- **Personalized Recommendations**: AI-driven suggestions for saving money +- **Tax Form Analysis**: Upload W2s, 1099s and get instant summaries -# Stop all services -docker-compose down -``` +### 💳 Smart Credit Card Optimization +- **Rewards Maximization**: Get the best card for each purchase category +- **Spending Pattern Analysis**: Identify where you're leaving money on the table +- **Multi-Card Strategy**: Optimize rewards across all your cards -Access at **http://localhost:3000** +### 📈 Investment Portfolio Management +- **Real-Time Performance**: Track returns with live Yahoo Finance data +- **Time-Based Analysis**: WoW, MoM, QoQ, YoY performance tracking +- **Tax Loss Harvesting**: Find opportunities to reduce your tax bill +- **Rebalancing Recommendations**: Compare to Bogleheads, Three-Fund Portfolio + +### 📋 Tax Document Processing +- **Supported Forms**: W2, 1099-INT, 1099-DIV, 1099-B, 1099-MISC, 1095-C +- **LLM-Powered Extraction**: Accurate parsing with structured output +- **Tax Estimate**: Calculate potential refund or amount owed +- **State Tax Support**: Extracts state wages and withholdings + +### 💰 Transaction & Spending Analysis +- **100+ Financial Institutions**: Connect via Plaid +- **Subscription Detection**: Find forgotten recurring charges +- **Category Analysis**: Understand where your money goes +- **Historical Comparisons**: Compare spending across time periods --- @@ -367,93 +194,68 @@ Access at **http://localhost:3000** ### Tech Stack -#### Frontend (`financegpt_web/`) -- **Framework**: Next.js 15 with App Router -- **Language**: TypeScript -- **Styling**: Tailwind CSS -- **UI Components**: Shadcn/ui, Radix UI -- **State Management**: Jotai -- **Animations**: Framer Motion -- **Real-Time**: ElectricSQL - -#### Backend (`financegpt_backend/`) -- **Framework**: FastAPI -- **Language**: Python 3.11+ -- **AI/ML**: LangChain, OpenAI GPT-4 -- **Database**: PostgreSQL with SQLAlchemy -- **Task Queue**: Celery with Redis -- **Financial Data**: Plaid API -- **Authentication**: OAuth 2.0 - -#### Infrastructure -- **Database**: PostgreSQL 15 -- **Cache**: Redis -- **Message Broker**: Redis (for Celery) -- **Container**: Docker & Docker Compose +| Layer | Technology | Purpose | +|-------|------------|---------| +| **Frontend** | Next.js 15, TypeScript, Tailwind | Modern web UI with server components | +| **Backend** | FastAPI, Python 3.11+ | Async API with auto-generated docs | +| **Database** | PostgreSQL + pgvector | Relational + vector search | +| **Task Queue** | Celery + Redis | Background document processing | +| **AI** | LiteLLM | Provider-agnostic (OpenAI, Anthropic, Ollama) | +| **Banking** | Plaid API | 100+ financial institution connections | +| **Auth** | Better Auth | OAuth 2.0, Google Sign-In | ### Project Structure ``` FinanceGPT/ -├── financegpt_web/ # Next.js frontend application +├── financegpt_web/ # Next.js frontend │ ├── app/ # App router pages │ ├── components/ # React components -│ ├── lib/ # Utilities and helpers -│ └── public/ # Static assets -├── financegpt_backend/ # FastAPI backend application +│ └── lib/ # Utilities +├── financegpt_backend/ # FastAPI backend │ ├── app/ │ │ ├── agents/ # AI agents and tools +│ │ ├── parsers/ # Tax form parsers │ │ ├── routes/ # API endpoints -│ │ ├── services/ # Business logic -│ │ ├── tasks/ # Celery tasks -│ │ └── utils/ # Utilities +│ │ └── tasks/ # Celery tasks │ └── alembic/ # Database migrations -├── financegpt_browser_extension/ # Browser extension -├── docker-compose.yml # Docker services configuration -└── README.md # This file +├── docker-compose.yml # Full stack deployment +├── docker-compose.quickstart.yml # All-in-one container +└── dev.sh # Local development script ``` --- -## 📖 Documentation - -### Configuration +## 📖 Configuration -#### Plaid API Setup -1. Sign up for a [Plaid account](https://plaid.com/) -2. Get your API keys (Client ID and Secret) -3. Add to `financegpt_backend/.env`: - ```env - PLAID_CLIENT_ID=your_client_id - PLAID_SECRET=your_secret - PLAID_ENV=sandbox # or development/production - ``` +### Required Environment Variables -#### OpenAI API Setup -1. Get your API key from [OpenAI](https://platform.openai.com/) -2. Add to `financegpt_backend/.env`: - ```env - OPENAI_API_KEY=your_api_key - ``` - -#### Database Configuration ```env -DATABASE_URL=postgresql://user:password@localhost:5432/financegpt +# LLM Provider (choose one) +OPENAI_API_KEY=sk-... +# or ANTHROPIC_API_KEY=sk-ant-... +# or GOOGLE_API_KEY=AIza... + +# Plaid (for bank connections) +PLAID_CLIENT_ID=your_client_id +PLAID_SECRET=your_secret +PLAID_ENV=sandbox + +# Security +SECRET_KEY=your-random-secret-key ``` -### API Endpoints +### Optional Configuration -#### Financial Data -- `GET /api/accounts` - List all connected accounts -- `GET /api/transactions` - Get transactions -- `POST /api/plaid/link-token` - Create Plaid Link token -- `POST /api/plaid/exchange-token` - Exchange public token +```env +# Document Processing +UNSTRUCTURED_API_KEY=... # For PDF parsing +ETL_SERVICE=DOCLING # Or UNSTRUCTURED -#### AI Features -- `POST /api/chat` - Chat with AI assistant -- `POST /api/analyze/spending` - Analyze spending patterns -- `POST /api/optimize/credit-card` - Get credit card recommendations -- `GET /api/insights` - Get personalized insights +# Voice Features +TTS_SERVICE=local/kokoro +STT_SERVICE=local/base +``` --- @@ -461,42 +263,19 @@ DATABASE_URL=postgresql://user:password@localhost:5432/financegpt ### Running Tests -**Frontend:** -```bash -cd financegpt_web -pnpm test -``` - -**Backend:** -```bash -cd financegpt_backend -pytest -``` - -### Code Quality - -**Frontend:** ```bash -pnpm lint -pnpm type-check -``` +# Backend +cd financegpt_backend && pytest -**Backend:** -```bash -ruff check . -mypy . +# Frontend +cd financegpt_web && pnpm test ``` ### Database Migrations -**Create a new migration:** ```bash cd financegpt_backend alembic revision --autogenerate -m "Description" -``` - -**Apply migrations:** -```bash alembic upgrade head ``` @@ -504,7 +283,7 @@ alembic upgrade head ## 🤝 Contributing -We welcome contributions! Please see our [Contributing Guide](CONTRIBUTING.md) for details. +We welcome contributions! Please see our [Contributing Guide](CONTRIBUTING.md). 1. Fork the repository 2. Create your feature branch (`git checkout -b feature/amazing-feature`) @@ -522,18 +301,16 @@ This project is licensed under the Apache License 2.0 - see the [LICENSE](LICENS ## 🙏 Acknowledgments -- Built with [Next.js](https://nextjs.org/), [FastAPI](https://fastapi.tiangolo.com/), and [LangChain](https://langchain.com/) +- Built on [SurfSense](https://github.com/MODSetter/SurfSense), an open-source NotebookLM alternative - Financial data powered by [Plaid](https://plaid.com/) -- AI capabilities powered by [OpenAI](https://openai.com/) +- AI capabilities via [LiteLLM](https://github.com/BerriAI/litellm) --- -## 📧 Contact +
-For questions or support, please open an issue or contact us at support@financegpt.com +Made with ❤️ for anyone who's ever stared at a W2 wondering what it all means. ---- +**[⬆ Back to Top](#financegpt-)** -
-Made with ❤️ by the FinanceGPT Team
diff --git a/dev.sh b/dev.sh new file mode 100755 index 0000000..5686066 --- /dev/null +++ b/dev.sh @@ -0,0 +1,69 @@ +#!/bin/bash +# FinanceGPT Local Development Script +# Runs all services in separate terminal tabs (macOS) + +set -e + +SCRIPT_DIR="$(cd "$(dirname "${BASH_SOURCE[0]}")" && pwd)" +BACKEND_DIR="$SCRIPT_DIR/financegpt_backend" +WEB_DIR="$SCRIPT_DIR/financegpt_web" + +echo "🚀 Starting FinanceGPT development environment..." +echo " Backend: $BACKEND_DIR" +echo " Web: $WEB_DIR" + +# Check if running on macOS +if [[ "$OSTYPE" != "darwin"* ]]; then + echo "❌ This script is designed for macOS. For Linux, use tmux or run manually." + echo "" + echo "Commands to run manually:" + echo " Terminal 1 (Backend API):" + echo " cd $BACKEND_DIR && source .venv/bin/activate && uv pip install -e . && python main.py --reload" + echo "" + echo " Terminal 2 (Celery Worker):" + echo " cd $BACKEND_DIR && source .venv/bin/activate && celery -A celery_worker.celery_app worker --loglevel=info --concurrency=1 --pool=solo" + echo "" + echo " Terminal 3 (Frontend):" + echo " cd $WEB_DIR && pnpm run dev" + exit 1 +fi + +# Create venv if it doesn't exist +if [ ! -d "$BACKEND_DIR/.venv" ]; then + echo "📦 Creating virtual environment..." + cd "$BACKEND_DIR" + python3 -m venv .venv +fi + +# Open Terminal tabs using AppleScript +osascript < None: + """Change tax_forms.document_id FK from SET NULL to CASCADE.""" + # Drop the existing foreign key constraint + op.drop_constraint('tax_forms_document_id_fkey', 'tax_forms', type_='foreignkey') + + # Recreate with CASCADE on delete + op.create_foreign_key( + 'tax_forms_document_id_fkey', + 'tax_forms', + 'documents', + ['document_id'], + ['id'], + ondelete='CASCADE' + ) + + +def downgrade() -> None: + """Revert to SET NULL behavior.""" + op.drop_constraint('tax_forms_document_id_fkey', 'tax_forms', type_='foreignkey') + + op.create_foreign_key( + 'tax_forms_document_id_fkey', + 'tax_forms', + 'documents', + ['document_id'], + ['id'], + ondelete='SET NULL' + ) diff --git a/financegpt_backend/app/agents/new_chat/system_prompt.py b/financegpt_backend/app/agents/new_chat/system_prompt.py index d96b625..a088979 100644 --- a/financegpt_backend/app/agents/new_chat/system_prompt.py +++ b/financegpt_backend/app/agents/new_chat/system_prompt.py @@ -519,14 +519,15 @@ 11. analyze_tax_data: Query uploaded and processed tax forms to answer tax questions. - **USE THIS TOOL** when users ask about: * Tax form data: "How much did I earn?", "What were my wages?" - * Tax withholdings: "How much federal tax was withheld?" + * Tax withholdings: "How much federal tax was withheld?", "State taxes withheld?" * Interest income: "Did I have interest income?", "1099-INT summary?" * Dividend income: "What dividends did I receive?" * Capital gains: "Stock sale gains/losses?", "1099-B summary?" - * W2 employment: "Where did I work?", "Wages by employer?" + * W2 employment: "Where did I work?", "Wages by employer?", "State wages?" * Tax year summaries: "2024 tax summary", "Total income 2024" + * Tax estimates/refunds: "Will I get a refund?", "How much do I owe?", "Estimate my taxes" - IMPORTANT: This tool queries ONLY uploaded tax forms (W2, 1099-MISC, 1099-INT, 1099-DIV, 1099-B) - - Does NOT calculate estimates or current year projections - only historical data from uploaded forms + - W2 forms include state tax info (state code, state wages, state income tax withheld) - Args: - query_type: Type of analysis (required). Options: * "income_summary": Total income across all sources @@ -534,12 +535,13 @@ * "interest_income": Interest from 1099-INT forms * "dividends_income": Dividends from 1099-DIV forms * "capital_gains": Capital gains from 1099-B forms - * "w2_summary": W2 employment wages and withholdings + * "w2_summary": W2 employment wages and withholdings (includes state tax info) * "all_forms": List all uploaded tax forms + * "tax_estimate": Estimate federal tax liability and potential refund/amount owed - tax_year: Specific tax year (e.g., 2024) or omit for all years - form_types: Optional filter by form types (e.g., ["W2", "1099-INT"]) - Returns: Structured tax data with totals, breakdowns, and per-form details - - Privacy: All PII (SSN, EIN) is hashed - never exposed in responses + - Privacy: All PII (SSN, EIN) is masked - never exposed in responses @@ -645,14 +647,25 @@ - User: "Which companies did I work for in 2024?" - Call: `analyze_tax_data(query_type="w2_summary", tax_year=2024)` - - Returns: W2 forms with employers, wages, and withholdings - - Summarize total wages and tax withheld + - Returns: W2 forms with employers, wages, withholdings, and state tax info + - Summarize total wages and federal/state tax withheld - User: "List all my uploaded tax forms" - Call: `analyze_tax_data(query_type="all_forms")` - Returns: All tax forms with types, years, and processing status - Note which forms need review (low confidence extractions) +- User: "Will I get a tax refund?" or "How much do I owe?" + - Call: `analyze_tax_data(query_type="tax_estimate", tax_year=2024)` + - Returns: Estimated federal tax, total withheld, refund or amount owed + - Note: Simplified estimate using single filer status and standard deduction + - Always recommend consulting a tax professional for accuracy + +- User: "What state income tax did I pay?" + - Call: `analyze_tax_data(query_type="w2_summary", tax_year=2024)` + - Returns: W2 data including state_code, state_wages, and state_income_tax withheld + - Summarize state taxes by state if multiple W2s from different states + - User: "How much more am I spending this month compared to last month?" - First call: `search_knowledge_base(query="transactions spending", start_date="2025-12-01", end_date="2025-12-31")` (Dec) - Second call: `search_knowledge_base(query="transactions spending", start_date="2026-01-01", end_date="2026-01-26")` (Jan) diff --git a/financegpt_backend/app/agents/new_chat/tools/tax_analysis.py b/financegpt_backend/app/agents/new_chat/tools/tax_analysis.py index 0f4413b..3e58169 100644 --- a/financegpt_backend/app/agents/new_chat/tools/tax_analysis.py +++ b/financegpt_backend/app/agents/new_chat/tools/tax_analysis.py @@ -5,22 +5,47 @@ - "What was my total federal tax withheld?" - "Did I have any interest income?" - "What were my capital gains from stock sales?" +- "Can you estimate my tax refund?" """ import logging from datetime import datetime from decimal import Decimal from typing import Any +from uuid import UUID from langchain_core.tools import tool from sqlalchemy import and_, desc, func, select from sqlalchemy.ext.asyncio import AsyncSession +from sqlalchemy.orm import selectinload -from app.schemas.tax_forms import TaxFormWithDetails +from app.db import ( + TaxForm, + W2Form, + Form1099Misc, + Form1099Int, + Form1099Div, + Form1099B, +) logger = logging.getLogger(__name__) +# 2024 Federal Tax Brackets (Single Filers) +TAX_BRACKETS_2024_SINGLE = [ + (11600, Decimal("0.10")), # 10% up to $11,600 + (47150, Decimal("0.12")), # 12% up to $47,150 + (100525, Decimal("0.22")), # 22% up to $100,525 + (191950, Decimal("0.24")), # 24% up to $191,950 + (243725, Decimal("0.32")), # 32% up to $243,725 + (609350, Decimal("0.35")), # 35% up to $609,350 + (float('inf'), Decimal("0.37")), # 37% above $609,350 +] + +# Standard Deduction 2024 +STANDARD_DEDUCTION_2024 = Decimal("14600") + + def create_tax_analysis_tool(user_id: str, search_space_id: int, db_session: AsyncSession): """Create the tax analysis tool for the agent. @@ -42,12 +67,19 @@ async def analyze_tax_data( """Query uploaded tax forms to answer tax-related questions. Use this tool when users ask about income, taxes withheld, interest, dividends, - capital gains, or W2 employment information from their uploaded tax documents. + capital gains, W2 employment information, or tax estimates/refunds from their + uploaded tax documents. Args: - query_type: Type of tax analysis - "income_summary", "tax_summary", - "interest_income", "dividends_income", "capital_gains", - "w2_summary", or "all_forms" + query_type: Type of tax analysis: + - "income_summary": Total income across all sources + - "tax_summary": Total taxes withheld from all sources + - "interest_income": Interest income from 1099-INT forms + - "dividends_income": Dividend income from 1099-DIV forms + - "capital_gains": Capital gains from 1099-B forms + - "w2_summary": W2 wage and withholding summary + - "all_forms": List all tax forms + - "tax_estimate": Estimate tax liability and potential refund tax_year: Specific tax year (e.g., 2024) or None for all years form_types: Optional list of form types to filter (e.g., ["W2", "1099-INT"]) @@ -55,6 +87,7 @@ async def analyze_tax_data( Dictionary with analysis results including totals, breakdowns, and details """ return await _analyze_tax_data_impl( + db_session=db_session, user_id=user_id, search_space_id=search_space_id, query_type=query_type, @@ -66,6 +99,7 @@ async def analyze_tax_data( async def _analyze_tax_data_impl( + db_session: AsyncSession, user_id: str, search_space_id: int, query_type: str, @@ -73,150 +107,662 @@ async def _analyze_tax_data_impl( form_types: list[str] | None = None, ) -> dict[str, Any]: """Implementation of tax data analysis.""" - # Note: Actual database queries would go here - # For now, returning placeholder structure if query_type == "income_summary": - return await _get_income_summary(user_id, search_space_id, tax_year) + return await _get_income_summary(db_session, user_id, search_space_id, tax_year) elif query_type == "tax_summary": - return await _get_tax_summary(user_id, search_space_id, tax_year) + return await _get_tax_summary(db_session, user_id, search_space_id, tax_year) elif query_type == "interest_income": - return await _get_interest_income(user_id, search_space_id, tax_year) + return await _get_interest_income(db_session, user_id, search_space_id, tax_year) elif query_type == "dividends_income": - return await _get_dividends_income(user_id, search_space_id, tax_year) + return await _get_dividends_income(db_session, user_id, search_space_id, tax_year) elif query_type == "capital_gains": - return await _get_capital_gains(user_id, search_space_id, tax_year) + return await _get_capital_gains(db_session, user_id, search_space_id, tax_year) elif query_type == "w2_summary": - return await _get_w2_summary(user_id, search_space_id, tax_year) + return await _get_w2_summary(db_session, user_id, search_space_id, tax_year) elif query_type == "all_forms": - return await _get_all_forms(user_id, search_space_id, tax_year, form_types) + return await _get_all_forms(db_session, user_id, search_space_id, tax_year, form_types) + elif query_type == "tax_estimate": + return await _get_tax_estimate(db_session, user_id, search_space_id, tax_year) else: return {"error": f"Unknown query type: {query_type}"} +async def _get_w2_summary( + db_session: AsyncSession, + user_id: str, + search_space_id: int, + tax_year: int | None, +) -> dict[str, Any]: + """Get W2 summary with actual database queries.""" + try: + # Build query for tax forms with W2 data + query = ( + select(TaxForm) + .options(selectinload(TaxForm.w2_form)) + .where( + and_( + TaxForm.user_id == UUID(user_id), + TaxForm.search_space_id == search_space_id, + TaxForm.form_type == "W2", + ) + ) + ) + + if tax_year: + query = query.where(TaxForm.tax_year == tax_year) + + query = query.order_by(TaxForm.tax_year.desc()) + + result = await db_session.execute(query) + tax_forms = result.scalars().all() + + if not tax_forms: + return { + "query_type": "w2_summary", + "tax_year": tax_year or "all years", + "has_data": False, + "message": "No W2 forms found. Upload your W2s to see employment income and withholdings.", + } + + # Aggregate data + employers = [] + total_wages = Decimal("0.00") + total_federal_withheld = Decimal("0.00") + total_ss_withheld = Decimal("0.00") + total_medicare_withheld = Decimal("0.00") + total_state_withheld = Decimal("0.00") + + for form in tax_forms: + w2 = form.w2_form + if w2: + employer_info = { + "employer_name": w2.employer_name or "Unknown Employer", + "tax_year": form.tax_year, + "wages": float(w2.wages_tips_compensation or 0), + "federal_withheld": float(w2.federal_income_tax_withheld or 0), + "social_security_withheld": float(w2.social_security_tax_withheld or 0), + "medicare_withheld": float(w2.medicare_tax_withheld or 0), + "state_withheld": float(w2.state_income_tax or 0), + "retirement_plan": w2.retirement_plan, + "processing_status": form.processing_status, + "needs_review": form.needs_review, + } + employers.append(employer_info) + + total_wages += w2.wages_tips_compensation or Decimal("0.00") + total_federal_withheld += w2.federal_income_tax_withheld or Decimal("0.00") + total_ss_withheld += w2.social_security_tax_withheld or Decimal("0.00") + total_medicare_withheld += w2.medicare_tax_withheld or Decimal("0.00") + total_state_withheld += w2.state_income_tax or Decimal("0.00") + + return { + "query_type": "w2_summary", + "tax_year": tax_year or "all years", + "has_data": True, + "employers": employers, + "total_wages": float(total_wages), + "total_federal_withheld": float(total_federal_withheld), + "total_social_security_withheld": float(total_ss_withheld), + "total_medicare_withheld": float(total_medicare_withheld), + "total_state_withheld": float(total_state_withheld), + "total_w2_forms": len(employers), + } + except Exception as e: + logger.exception("Error getting W2 summary") + return { + "query_type": "w2_summary", + "error": str(e), + } + + async def _get_income_summary( + db_session: AsyncSession, user_id: str, search_space_id: int, tax_year: int | None, ) -> dict[str, Any]: """Get total income across all sources.""" - # TODO: Implement actual database queries - # This would query W2s and 1099s to sum total income - return { - "query_type": "income_summary", - "tax_year": tax_year or "all years", - "total_w2_wages": Decimal("0.00"), - "total_1099_misc_income": Decimal("0.00"), - "total_interest_income": Decimal("0.00"), - "total_dividend_income": Decimal("0.00"), - "total_capital_gains": Decimal("0.00"), - "grand_total_income": Decimal("0.00"), - "message": "No tax forms uploaded yet. Please upload your W2 and 1099 forms to see income summary.", - } + try: + w2_data = await _get_w2_summary(db_session, user_id, search_space_id, tax_year) + interest_data = await _get_interest_income(db_session, user_id, search_space_id, tax_year) + dividend_data = await _get_dividends_income(db_session, user_id, search_space_id, tax_year) + capital_gains_data = await _get_capital_gains(db_session, user_id, search_space_id, tax_year) + misc_data = await _get_misc_income(db_session, user_id, search_space_id, tax_year) + + total_w2_wages = w2_data.get("total_wages", 0) if w2_data.get("has_data") else 0 + total_interest = interest_data.get("total_interest", 0) if interest_data.get("has_data") else 0 + total_dividends = dividend_data.get("total_ordinary_dividends", 0) if dividend_data.get("has_data") else 0 + total_capital_gains = capital_gains_data.get("net_capital_gains", 0) if capital_gains_data.get("has_data") else 0 + total_misc = misc_data.get("total_misc_income", 0) if misc_data.get("has_data") else 0 + + grand_total = total_w2_wages + total_interest + total_dividends + total_capital_gains + total_misc + + has_any_data = any([ + w2_data.get("has_data"), + interest_data.get("has_data"), + dividend_data.get("has_data"), + capital_gains_data.get("has_data"), + misc_data.get("has_data"), + ]) + + if not has_any_data: + return { + "query_type": "income_summary", + "tax_year": tax_year or "all years", + "has_data": False, + "message": "No tax forms uploaded yet. Please upload your W2 and 1099 forms to see income summary.", + } + + return { + "query_type": "income_summary", + "tax_year": tax_year or "all years", + "has_data": True, + "total_w2_wages": total_w2_wages, + "total_interest_income": total_interest, + "total_dividend_income": total_dividends, + "total_capital_gains": total_capital_gains, + "total_misc_income": total_misc, + "grand_total_income": grand_total, + "breakdown": { + "w2": w2_data, + "interest": interest_data, + "dividends": dividend_data, + "capital_gains": capital_gains_data, + "misc": misc_data, + }, + } + except Exception as e: + logger.exception("Error getting income summary") + return {"query_type": "income_summary", "error": str(e)} async def _get_tax_summary( + db_session: AsyncSession, user_id: str, search_space_id: int, tax_year: int | None, ) -> dict[str, Any]: """Get total taxes withheld across all sources.""" - # TODO: Implement actual database queries - return { - "query_type": "tax_summary", - "tax_year": tax_year or "all years", - "total_federal_withheld": Decimal("0.00"), - "total_social_security_withheld": Decimal("0.00"), - "total_medicare_withheld": Decimal("0.00"), - "total_state_withheld": Decimal("0.00"), - "grand_total_withheld": Decimal("0.00"), - "message": "No tax forms uploaded yet. Please upload your W2 and 1099 forms to see tax withholdings.", - } + try: + w2_data = await _get_w2_summary(db_session, user_id, search_space_id, tax_year) + interest_data = await _get_interest_income(db_session, user_id, search_space_id, tax_year) + dividend_data = await _get_dividends_income(db_session, user_id, search_space_id, tax_year) + + total_federal = w2_data.get("total_federal_withheld", 0) if w2_data.get("has_data") else 0 + total_federal += interest_data.get("total_federal_withheld", 0) if interest_data.get("has_data") else 0 + total_federal += dividend_data.get("total_federal_withheld", 0) if dividend_data.get("has_data") else 0 + + total_ss = w2_data.get("total_social_security_withheld", 0) if w2_data.get("has_data") else 0 + total_medicare = w2_data.get("total_medicare_withheld", 0) if w2_data.get("has_data") else 0 + total_state = w2_data.get("total_state_withheld", 0) if w2_data.get("has_data") else 0 + + has_data = w2_data.get("has_data") or interest_data.get("has_data") or dividend_data.get("has_data") + + if not has_data: + return { + "query_type": "tax_summary", + "tax_year": tax_year or "all years", + "has_data": False, + "message": "No tax forms uploaded yet. Please upload your W2 and 1099 forms to see tax withholdings.", + } + + return { + "query_type": "tax_summary", + "tax_year": tax_year or "all years", + "has_data": True, + "total_federal_withheld": total_federal, + "total_social_security_withheld": total_ss, + "total_medicare_withheld": total_medicare, + "total_state_withheld": total_state, + "grand_total_withheld": total_federal + total_ss + total_medicare + total_state, + } + except Exception as e: + logger.exception("Error getting tax summary") + return {"query_type": "tax_summary", "error": str(e)} async def _get_interest_income( + db_session: AsyncSession, user_id: str, search_space_id: int, tax_year: int | None, ) -> dict[str, Any]: """Get interest income from 1099-INT forms.""" - # TODO: Implement actual database queries - return { - "query_type": "interest_income", - "tax_year": tax_year or "all years", - "total_interest": Decimal("0.00"), - "sources": [], - "message": "No 1099-INT forms found. Upload your interest income statements to see details.", - } + try: + query = ( + select(TaxForm) + .options(selectinload(TaxForm.form_1099_int)) + .where( + and_( + TaxForm.user_id == UUID(user_id), + TaxForm.search_space_id == search_space_id, + TaxForm.form_type == "1099-INT", + ) + ) + ) + + if tax_year: + query = query.where(TaxForm.tax_year == tax_year) + + result = await db_session.execute(query) + tax_forms = result.scalars().all() + + if not tax_forms: + return { + "query_type": "interest_income", + "tax_year": tax_year or "all years", + "has_data": False, + "message": "No 1099-INT forms found. Upload your interest income statements to see details.", + } + + sources = [] + total_interest = Decimal("0.00") + total_federal_withheld = Decimal("0.00") + + for form in tax_forms: + int_form = form.form_1099_int + if int_form: + source_info = { + "payer_name": int_form.payer_name or "Unknown Payer", + "tax_year": form.tax_year, + "interest_income": float(int_form.interest_income or 0), + "federal_withheld": float(int_form.federal_income_tax_withheld or 0), + } + sources.append(source_info) + + total_interest += int_form.interest_income or Decimal("0.00") + total_federal_withheld += int_form.federal_income_tax_withheld or Decimal("0.00") + + return { + "query_type": "interest_income", + "tax_year": tax_year or "all years", + "has_data": True, + "total_interest": float(total_interest), + "total_federal_withheld": float(total_federal_withheld), + "sources": sources, + } + except Exception as e: + logger.exception("Error getting interest income") + return {"query_type": "interest_income", "error": str(e)} async def _get_dividends_income( + db_session: AsyncSession, user_id: str, search_space_id: int, tax_year: int | None, ) -> dict[str, Any]: """Get dividend income from 1099-DIV forms.""" - # TODO: Implement actual database queries - return { - "query_type": "dividends_income", - "tax_year": tax_year or "all years", - "total_ordinary_dividends": Decimal("0.00"), - "total_qualified_dividends": Decimal("0.00"), - "sources": [], - "message": "No 1099-DIV forms found. Upload your dividend income statements to see details.", - } + try: + query = ( + select(TaxForm) + .options(selectinload(TaxForm.form_1099_div)) + .where( + and_( + TaxForm.user_id == UUID(user_id), + TaxForm.search_space_id == search_space_id, + TaxForm.form_type == "1099-DIV", + ) + ) + ) + + if tax_year: + query = query.where(TaxForm.tax_year == tax_year) + + result = await db_session.execute(query) + tax_forms = result.scalars().all() + + if not tax_forms: + return { + "query_type": "dividends_income", + "tax_year": tax_year or "all years", + "has_data": False, + "message": "No 1099-DIV forms found. Upload your dividend income statements to see details.", + } + + sources = [] + total_ordinary = Decimal("0.00") + total_qualified = Decimal("0.00") + total_capital_gain_dist = Decimal("0.00") + total_federal_withheld = Decimal("0.00") + + for form in tax_forms: + div_form = form.form_1099_div + if div_form: + source_info = { + "payer_name": div_form.payer_name or "Unknown Payer", + "tax_year": form.tax_year, + "ordinary_dividends": float(div_form.total_ordinary_dividends or 0), + "qualified_dividends": float(div_form.qualified_dividends or 0), + "capital_gain_distributions": float(div_form.total_capital_gain_distributions or 0), + "federal_withheld": float(div_form.federal_income_tax_withheld or 0), + } + sources.append(source_info) + + total_ordinary += div_form.total_ordinary_dividends or Decimal("0.00") + total_qualified += div_form.qualified_dividends or Decimal("0.00") + total_capital_gain_dist += div_form.total_capital_gain_distributions or Decimal("0.00") + total_federal_withheld += div_form.federal_income_tax_withheld or Decimal("0.00") + + return { + "query_type": "dividends_income", + "tax_year": tax_year or "all years", + "has_data": True, + "total_ordinary_dividends": float(total_ordinary), + "total_qualified_dividends": float(total_qualified), + "total_capital_gain_distributions": float(total_capital_gain_dist), + "total_federal_withheld": float(total_federal_withheld), + "sources": sources, + } + except Exception as e: + logger.exception("Error getting dividends income") + return {"query_type": "dividends_income", "error": str(e)} async def _get_capital_gains( + db_session: AsyncSession, user_id: str, search_space_id: int, tax_year: int | None, ) -> dict[str, Any]: """Get capital gains from 1099-B forms.""" - # TODO: Implement actual database queries - return { - "query_type": "capital_gains", - "tax_year": tax_year or "all years", - "total_short_term_gains": Decimal("0.00"), - "total_long_term_gains": Decimal("0.00"), - "total_realized_gains": Decimal("0.00"), - "transactions": [], - "message": "No 1099-B forms found. Upload your brokerage statements to see capital gains.", - } + try: + query = ( + select(TaxForm) + .options(selectinload(TaxForm.form_1099_b)) + .where( + and_( + TaxForm.user_id == UUID(user_id), + TaxForm.search_space_id == search_space_id, + TaxForm.form_type == "1099-B", + ) + ) + ) + + if tax_year: + query = query.where(TaxForm.tax_year == tax_year) + + result = await db_session.execute(query) + tax_forms = result.scalars().all() + + if not tax_forms: + return { + "query_type": "capital_gains", + "tax_year": tax_year or "all years", + "has_data": False, + "message": "No 1099-B forms found. Upload your brokerage statements to see capital gains.", + } + + sources = [] + total_proceeds = Decimal("0.00") + total_cost_basis = Decimal("0.00") + + for form in tax_forms: + b_form = form.form_1099_b + if b_form: + proceeds = b_form.proceeds or Decimal("0.00") + cost_basis = b_form.cost_basis or Decimal("0.00") + gain_loss = proceeds - cost_basis + + source_info = { + "broker_name": b_form.payer_name or "Unknown Broker", + "tax_year": form.tax_year, + "proceeds": float(proceeds), + "cost_basis": float(cost_basis), + "gain_loss": float(gain_loss), + } + sources.append(source_info) + + total_proceeds += proceeds + total_cost_basis += cost_basis + + net_gains = total_proceeds - total_cost_basis + + return { + "query_type": "capital_gains", + "tax_year": tax_year or "all years", + "has_data": True, + "total_proceeds": float(total_proceeds), + "total_cost_basis": float(total_cost_basis), + "net_capital_gains": float(net_gains), + "sources": sources, + } + except Exception as e: + logger.exception("Error getting capital gains") + return {"query_type": "capital_gains", "error": str(e)} -async def _get_w2_summary( +async def _get_misc_income( + db_session: AsyncSession, user_id: str, search_space_id: int, tax_year: int | None, ) -> dict[str, Any]: - """Get W2 summary.""" - # TODO: Implement actual database queries - return { - "query_type": "w2_summary", - "tax_year": tax_year or "all years", - "employers": [], - "total_wages": Decimal("0.00"), - "total_federal_withheld": Decimal("0.00"), - "total_social_security_withheld": Decimal("0.00"), - "total_medicare_withheld": Decimal("0.00"), - "message": "No W2 forms found. Upload your W2s to see employment income and withholdings.", - } + """Get miscellaneous income from 1099-MISC forms.""" + try: + query = ( + select(TaxForm) + .options(selectinload(TaxForm.form_1099_misc)) + .where( + and_( + TaxForm.user_id == UUID(user_id), + TaxForm.search_space_id == search_space_id, + TaxForm.form_type == "1099-MISC", + ) + ) + ) + + if tax_year: + query = query.where(TaxForm.tax_year == tax_year) + + result = await db_session.execute(query) + tax_forms = result.scalars().all() + + if not tax_forms: + return { + "query_type": "misc_income", + "tax_year": tax_year or "all years", + "has_data": False, + } + + sources = [] + total_misc = Decimal("0.00") + + for form in tax_forms: + misc_form = form.form_1099_misc + if misc_form: + misc_income = ( + (misc_form.rents or Decimal("0.00")) + + (misc_form.royalties or Decimal("0.00")) + + (misc_form.other_income or Decimal("0.00")) + ) + + source_info = { + "payer_name": misc_form.payer_name or "Unknown Payer", + "tax_year": form.tax_year, + "rents": float(misc_form.rents or 0), + "royalties": float(misc_form.royalties or 0), + "other_income": float(misc_form.other_income or 0), + "total": float(misc_income), + } + sources.append(source_info) + total_misc += misc_income + + return { + "query_type": "misc_income", + "tax_year": tax_year or "all years", + "has_data": True, + "total_misc_income": float(total_misc), + "sources": sources, + } + except Exception as e: + logger.exception("Error getting misc income") + return {"query_type": "misc_income", "error": str(e)} async def _get_all_forms( + db_session: AsyncSession, user_id: str, search_space_id: int, tax_year: int | None, form_types: list[str] | None, ) -> dict[str, Any]: """Get all tax forms with optional filters.""" - # TODO: Implement actual database queries - return { - "query_type": "all_forms", - "tax_year": tax_year or "all years", - "form_types_filter": form_types, - "forms": [], - "total_forms": 0, - "message": "No tax forms uploaded yet. Upload W2s and 1099s to get started.", - } + try: + query = ( + select(TaxForm) + .where( + and_( + TaxForm.user_id == UUID(user_id), + TaxForm.search_space_id == search_space_id, + ) + ) + .order_by(TaxForm.tax_year.desc(), TaxForm.form_type) + ) + + if tax_year: + query = query.where(TaxForm.tax_year == tax_year) + + if form_types: + query = query.where(TaxForm.form_type.in_(form_types)) + + result = await db_session.execute(query) + tax_forms = result.scalars().all() + + if not tax_forms: + return { + "query_type": "all_forms", + "tax_year": tax_year or "all years", + "form_types_filter": form_types, + "has_data": False, + "total_forms": 0, + "message": "No tax forms uploaded yet. Upload W2s and 1099s to get started.", + } + + forms_list = [ + { + "id": str(form.id), + "form_type": form.form_type, + "tax_year": form.tax_year, + "processing_status": form.processing_status, + "extraction_method": form.extraction_method, + "needs_review": form.needs_review, + "uploaded_at": form.uploaded_at.isoformat() if form.uploaded_at else None, + } + for form in tax_forms + ] + + return { + "query_type": "all_forms", + "tax_year": tax_year or "all years", + "form_types_filter": form_types, + "has_data": True, + "forms": forms_list, + "total_forms": len(forms_list), + } + except Exception as e: + logger.exception("Error getting all forms") + return {"query_type": "all_forms", "error": str(e)} + + +async def _get_tax_estimate( + db_session: AsyncSession, + user_id: str, + search_space_id: int, + tax_year: int | None, +) -> dict[str, Any]: + """Estimate tax liability and potential refund. + + This provides a simplified estimate based on single filer status. + For accurate tax calculations, users should consult a tax professional. + """ + try: + # Get income summary first + income_data = await _get_income_summary(db_session, user_id, search_space_id, tax_year) + + if not income_data.get("has_data"): + return { + "query_type": "tax_estimate", + "tax_year": tax_year or "all years", + "has_data": False, + "message": "No tax forms uploaded yet. Please upload your W2 and 1099 forms to get a tax estimate.", + } + + # Get withholdings + tax_data = await _get_tax_summary(db_session, user_id, search_space_id, tax_year) + + # Calculate estimated tax + gross_income = Decimal(str(income_data.get("grand_total_income", 0))) + + # Apply standard deduction + taxable_income = max(Decimal("0.00"), gross_income - STANDARD_DEDUCTION_2024) + + # Calculate tax using brackets + estimated_tax = _calculate_tax_from_brackets(taxable_income) + + # Get total withholdings + total_withheld = Decimal(str(tax_data.get("total_federal_withheld", 0))) + + # Calculate refund/owed + difference = total_withheld - estimated_tax + + if difference > 0: + refund_or_owed = "refund" + amount = float(difference) + else: + refund_or_owed = "owed" + amount = float(abs(difference)) + + return { + "query_type": "tax_estimate", + "tax_year": tax_year or 2024, + "has_data": True, + "disclaimer": "This is a simplified estimate assuming single filer status with standard deduction. For accurate calculations, please consult a tax professional.", + "gross_income": float(gross_income), + "standard_deduction": float(STANDARD_DEDUCTION_2024), + "taxable_income": float(taxable_income), + "estimated_federal_tax": float(estimated_tax), + "total_federal_withheld": float(total_withheld), + "estimate_result": refund_or_owed, + "estimate_amount": amount, + "breakdown": { + "income": { + "w2_wages": income_data.get("total_w2_wages", 0), + "interest": income_data.get("total_interest_income", 0), + "dividends": income_data.get("total_dividend_income", 0), + "capital_gains": income_data.get("total_capital_gains", 0), + "misc_income": income_data.get("total_misc_income", 0), + }, + "withholdings": { + "federal": tax_data.get("total_federal_withheld", 0), + "social_security": tax_data.get("total_social_security_withheld", 0), + "medicare": tax_data.get("total_medicare_withheld", 0), + "state": tax_data.get("total_state_withheld", 0), + }, + }, + } + except Exception as e: + logger.exception("Error calculating tax estimate") + return {"query_type": "tax_estimate", "error": str(e)} + + +def _calculate_tax_from_brackets(taxable_income: Decimal) -> Decimal: + """Calculate federal tax using progressive brackets.""" + if taxable_income <= 0: + return Decimal("0.00") + + tax = Decimal("0.00") + prev_bracket = Decimal("0") + + for bracket_max, rate in TAX_BRACKETS_2024_SINGLE: + bracket_max_decimal = Decimal(str(bracket_max)) + + if taxable_income <= bracket_max_decimal: + tax += (taxable_income - prev_bracket) * rate + break + else: + tax += (bracket_max_decimal - prev_bracket) * rate + prev_bracket = bracket_max_decimal + + return tax.quantize(Decimal("0.01")) diff --git a/financegpt_backend/app/celery_app.py b/financegpt_backend/app/celery_app.py index 36b0b64..f711d13 100644 --- a/financegpt_backend/app/celery_app.py +++ b/financegpt_backend/app/celery_app.py @@ -64,6 +64,7 @@ def parse_schedule_interval(interval: str) -> dict: "app.tasks.celery_tasks.schedule_checker_task", "app.tasks.celery_tasks.blocknote_migration_tasks", "app.tasks.celery_tasks.document_reindex_tasks", + "app.tasks.celery_tasks.tax_form_tasks", ], ) diff --git a/financegpt_backend/app/db.py b/financegpt_backend/app/db.py index 3932bf1..5688f05 100644 --- a/financegpt_backend/app/db.py +++ b/financegpt_backend/app/db.py @@ -981,7 +981,7 @@ class TaxForm(BaseModel): form_type = Column(String(20), nullable=False, index=True) # W2, 1099-MISC, 1099-INT, etc. tax_year = Column(Integer, nullable=False, index=True) document_id = Column( - Integer, ForeignKey("documents.id", ondelete="SET NULL"), nullable=True + Integer, ForeignKey("documents.id", ondelete="CASCADE"), nullable=True ) uploaded_at = Column(TIMESTAMP(timezone=True), server_default=text('now()'), nullable=False) processed_at = Column(TIMESTAMP(timezone=True), nullable=True) diff --git a/financegpt_backend/app/parsers/llm_csv_parser.py b/financegpt_backend/app/parsers/llm_csv_parser.py index 563752d..6136c8d 100644 --- a/financegpt_backend/app/parsers/llm_csv_parser.py +++ b/financegpt_backend/app/parsers/llm_csv_parser.py @@ -32,6 +32,80 @@ def __init__(self): """Initialize LLM CSV parser.""" super().__init__("LLM CSV Parser") + def _skip_preamble_rows(self, text_content: str) -> str: + """ + Skip preamble/header rows that appear before actual CSV data. + + Many brokerage CSV exports include metadata rows before the actual CSV headers. + Examples: + - Schwab: "Positions for account ****9247 as of 01/29/2026 04:00 PM ET" + - Fidelity: "Brokerage" or account name rows + - E*TRADE: Account summary rows + + Detection strategy: + 1. Find the first line that looks like a proper CSV header row + 2. A header row has multiple comma-separated fields (>=3) + 3. Subsequent data rows should have the same number of fields + 4. Preamble rows typically have fewer fields or are single values + + Args: + text_content: Raw CSV text content + + Returns: + CSV text with preamble rows removed + """ + lines = text_content.strip().split('\n') + if not lines: + return text_content + + def count_csv_fields(line: str) -> int: + """Count fields in a CSV line, handling quoted values with commas.""" + try: + # Use csv module to properly parse the line + reader = csv.reader(io.StringIO(line)) + fields = next(reader, []) + return len(fields) + except Exception: + # Fallback: simple comma count (less accurate but safe) + return line.count(',') + 1 + + def is_likely_header_row(line: str, next_lines: list[str]) -> bool: + """Check if this line looks like a CSV header row.""" + if not line.strip(): + return False + + field_count = count_csv_fields(line) + + # Headers should have at least 3 fields + if field_count < 3: + return False + + # Check if the next few non-empty lines have the same field count + # This indicates we found the actual CSV structure + matching_lines = 0 + for next_line in next_lines[:5]: # Check up to 5 subsequent lines + if not next_line.strip(): + continue + if count_csv_fields(next_line) == field_count: + matching_lines += 1 + + # If at least 2 subsequent lines match, this is likely the header + return matching_lines >= 2 + + # Find the first line that looks like a CSV header + for i, line in enumerate(lines): + if not line.strip(): + continue + + remaining_lines = lines[i+1:] + if is_likely_header_row(line, remaining_lines): + if i > 0: + logger.info(f"Skipped {i} preamble row(s) before CSV headers") + return '\n'.join(lines[i:]) + + # No preamble detected, return original content + return text_content + async def parse_file( self, file_content: bytes, @@ -56,6 +130,11 @@ async def parse_file( try: # Decode CSV text_content = file_content.decode("utf-8-sig") + + # Handle CSV files with preamble/header rows before actual CSV data + # (e.g., Schwab files start with "Positions for account ****XXXX as of...") + text_content = self._skip_preamble_rows(text_content) + csv_reader = csv.DictReader(io.StringIO(text_content)) # Get headers and first few rows for LLM analysis @@ -316,14 +395,18 @@ def safe_decimal(value: str | None, allow_negative: bool = False) -> Decimal | N holdings = [] + # Log schema for debugging + logger.debug(f"Applying holdings schema: {schema}") + for row in rows: try: # Extract symbol symbol_col = schema.get("symbol", {}).get("column") if not symbol_col or symbol_col not in row: + logger.debug(f"Symbol column '{symbol_col}' not found in row keys: {list(row.keys())}") continue symbol = str(row[symbol_col]).strip().upper() - if not symbol or symbol in ["", "N/A", "Total", "TOTAL"]: + if not symbol or symbol in ["", "N/A", "Total", "TOTAL", "ACCOUNT TOTAL", "CASH & CASH INVESTMENTS"]: continue # Extract quantity using safe_decimal diff --git a/financegpt_backend/app/parsers/tax_form_parser.py b/financegpt_backend/app/parsers/tax_form_parser.py index 57f9fcd..e3a4bde 100644 --- a/financegpt_backend/app/parsers/tax_form_parser.py +++ b/financegpt_backend/app/parsers/tax_form_parser.py @@ -9,20 +9,76 @@ Each tier returns confidence scores. If confidence < 0.85, escalate to next tier. """ +import decimal +import json import logging import re from decimal import Decimal from pathlib import Path from typing import Any, Literal +import litellm import pdfplumber +from pydantic import BaseModel, Field from unstructured.partition.pdf import partition_pdf -from app.utils.pii_masking import mask_tax_form_for_llm, validate_confidence_threshold +from app.utils.pii_masking import mask_tax_form_for_llm, validate_confidence_threshold, mask_pii_in_text, recover_pii_from_mapping logger = logging.getLogger(__name__) +# Pydantic models for structured LLM extraction +class W2ExtractedData(BaseModel): + """Structured W2 data extracted by LLM.""" + wages_tips_compensation: float | None = Field(None, description="Box 1: Wages, tips, other compensation") + federal_income_tax_withheld: float | None = Field(None, description="Box 2: Federal income tax withheld") + social_security_wages: float | None = Field(None, description="Box 3: Social security wages") + social_security_tax_withheld: float | None = Field(None, description="Box 4: Social security tax withheld") + medicare_wages: float | None = Field(None, description="Box 5: Medicare wages and tips") + medicare_tax_withheld: float | None = Field(None, description="Box 6: Medicare tax withheld") + # State tax fields + state_code: str | None = Field(None, description="Box 15: State code (e.g., CA, NY)") + state_wages: float | None = Field(None, description="Box 16: State wages, tips, etc.") + state_income_tax: float | None = Field(None, description="Box 17: State income tax withheld") + # Note: SSN/EIN extracted from heuristics, not LLM (privacy) + employer_name: str | None = Field(None, description="Employer's name") + retirement_plan: bool = Field(False, description="Box 13: Retirement plan checkbox marked") + + +class Form1099ExtractedData(BaseModel): + """Structured 1099 data extracted by LLM.""" + interest_income: float | None = Field(None, description="Interest income (1099-INT Box 1)") + dividend_income: float | None = Field(None, description="Total ordinary dividends (1099-DIV Box 1a)") + qualified_dividends: float | None = Field(None, description="Qualified dividends (1099-DIV Box 1b)") + capital_gain_distributions: float | None = Field(None, description="Capital gain distributions (1099-DIV Box 2a)") + proceeds: float | None = Field(None, description="Proceeds from sales (1099-B Box 1d)") + cost_basis: float | None = Field(None, description="Cost or other basis (1099-B Box 1e)") + rents: float | None = Field(None, description="Rents (1099-MISC Box 1)") + other_income: float | None = Field(None, description="Other income (1099-MISC Box 3)") + federal_tax_withheld: float | None = Field(None, description="Federal income tax withheld") + payer_name: str | None = Field(None, description="Payer's name") + # Note: Payer TIN extracted from heuristics, not LLM (privacy) + + +class Form1095CExtractedData(BaseModel): + """Structured 1095-C data extracted by LLM (Employer-Provided Health Insurance Offer and Coverage).""" + # Part I - Employee Information + employer_name: str | None = Field(None, description="Part I Line 1: Employer name") + employer_contact_phone: str | None = Field(None, description="Part I Line 9: Employer contact phone") + + # Part II - Employee Offer and Coverage (Line 14-16 codes for each month) + offer_of_coverage_code: str | None = Field(None, description="Line 14: Offer of Coverage code (e.g., 1A, 1B, 1C, 1E, 1H)") + employee_share_lowest_cost: float | None = Field(None, description="Line 15: Employee share of lowest cost monthly premium") + safe_harbor_code: str | None = Field(None, description="Line 16: Safe Harbor code (e.g., 2C, 2D, 2F)") + + # Coverage months (Part II Line 14 - which months had coverage) + covered_all_12_months: bool = Field(False, description="Whether coverage was offered for all 12 months") + covered_months: list[str] | None = Field(None, description="List of months with coverage if not all 12") + + # Part III - Covered Individuals (if applicable) + covered_individuals_count: int | None = Field(None, description="Number of covered individuals listed in Part III") + + class TaxFormParser: """Hybrid tax form parser with tiered extraction.""" @@ -41,11 +97,71 @@ def __init__(self): """Initialize parser.""" self.extraction_history: list[dict[str, Any]] = [] + async def parse_from_text( + self, + text: str, + form_type: Literal["W2", "1099-MISC", "1099-INT", "1099-DIV", "1099-B", "1095-C"], + tax_year: int, + llm_model: str | None = None, + ) -> dict[str, Any]: + """Parse tax form from pre-extracted text using LLM. + + This method is used when text has already been extracted from the PDF + (e.g., by Unstructured or Docling) and we just need to parse the fields. + + Always uses LLM extraction for accuracy. + + Args: + text: Pre-extracted text content + form_type: Type of tax form + tax_year: Tax year for the form + llm_model: Optional LLM model string (e.g., "gemini/gemini-2.0-flash") + If not provided, uses default model + + Returns: + Dictionary containing parsed data and confidence scores. + """ + logger.info(f"Parsing {form_type} from pre-extracted text ({len(text)} chars) using LLM") + + # Use LLM extraction for accuracy + extracted_data = await self._extract_with_llm(text, form_type, tax_year, llm_model=llm_model) + extraction_method = "llm_assisted" + + # If LLM extraction failed, fall back to heuristic parsing + if not extracted_data: + logger.warning(f"LLM extraction failed for {form_type}, falling back to heuristic parsing") + if form_type == "W2": + extracted_data = self._parse_w2_text(text) + elif form_type == "1099-MISC": + extracted_data = self._parse_1099_misc_text(text) + elif form_type == "1099-INT": + extracted_data = self._parse_1099_int_text(text) + elif form_type == "1099-DIV": + extracted_data = self._parse_1099_div_text(text) + elif form_type == "1099-B": + extracted_data = self._parse_1099_b_text(text) + elif form_type == "1095-C": + extracted_data = self._parse_1095_c_text(text) + else: + raise ValueError(f"Unsupported form type: {form_type}") + extraction_method = "heuristic" + + confidence_scores = self._calculate_confidence_scores(extracted_data) + + return { + "extracted_data": extracted_data, + "confidence_scores": confidence_scores, + "extraction_method": extraction_method, + "needs_review": not self._meets_confidence_threshold(confidence_scores), + "raw_extraction_data": {"text_length": len(text)}, + } + async def parse_tax_form( self, file_path: str | Path, - form_type: Literal["W2", "1099-MISC", "1099-INT", "1099-DIV", "1099-B"], + form_type: Literal["W2", "1099-MISC", "1099-INT", "1099-DIV", "1099-B", "1095-C"], tax_year: int, + llm_model: str | None = None, ) -> dict[str, Any]: """Parse tax form using tiered extraction strategy. @@ -53,6 +169,7 @@ async def parse_tax_form( file_path: Path to PDF file form_type: Type of tax form tax_year: Tax year for the form + llm_model: Optional LLM model string for fallback extraction Returns: Dictionary containing: @@ -249,13 +366,6 @@ async def _extract_llm_assisted( Best for: Verification of low-confidence fields, unusual layouts. IMPORTANT: All PII is masked before sending to LLM. """ - # TODO: Implement LLM-assisted extraction using instructor - # This would: - # 1. Convert PDF to image or extract text - # 2. Mask any PII found in previous extraction - # 3. Send to LLM with structured output schema - # 4. Return verified/extracted data - logger.warning("LLM-assisted extraction not yet implemented") # For now, return previous result marked as needs_review @@ -272,39 +382,217 @@ async def _extract_llm_assisted( "raw_extraction_data": {"status": "not_implemented"}, } + async def _extract_with_llm( + self, + text: str, + form_type: str, + tax_year: int, + llm_model: str | None = None, + ) -> dict[str, Any]: + """Use LLM to extract structured data from tax form text. + + This is called when heuristic parsing fails to find dollar amounts. + PII (SSN, EIN) is masked before sending to the LLM. + + Args: + text: The document text to parse + form_type: Type of tax form + tax_year: Tax year + llm_model: Optional LLM model string. If not provided, uses default. + """ + # Use provided model or fall back to default + model = llm_model or "gemini/gemini-2.0-flash" + logger.info(f"Using LLM extraction for {form_type} (tax year {tax_year}) with model: {model}") + + # Mask PII before sending to LLM + masked_text, pii_mapping = mask_pii_in_text(text) + if pii_mapping: + logger.info(f"Masked {len(pii_mapping)} PII items before LLM call") + + # Select the appropriate prompt based on form type + if form_type == "W2": + system_prompt = """You are a tax document parser. Extract W-2 form data from the provided text. + +Extract these specific fields: +- wages_tips_compensation: Box 1 - Wages, tips, other compensation (the main income amount) +- federal_income_tax_withheld: Box 2 - Federal income tax withheld +- social_security_wages: Box 3 - Social security wages (often same as Box 1, capped at ~$176,100 for 2025) +- social_security_tax_withheld: Box 4 - Social security tax withheld (about 6.2% of Box 3) +- medicare_wages: Box 5 - Medicare wages and tips (often same as Box 1) +- medicare_tax_withheld: Box 6 - Medicare tax withheld (about 1.45% of Box 5) +- state_code: Box 15 - Two-letter state code (e.g., CA, NY, TX) +- state_wages: Box 16 - State wages, tips, etc. +- state_income_tax: Box 17 - State income tax withheld +- employer_name: Name of the employer +- retirement_plan: True if Box 13 retirement plan is checked + +Note: SSN and EIN have been masked for privacy (shown as [SSN:***-**-XXXX] or [EIN:XXXXXXXX]). +Do NOT try to extract SSN or EIN values. + +Important: +- Look for dollar amounts formatted as $XX,XXX.XX or XX,XXX.XX +- The wages amount is typically the largest number (often $50,000 - $500,000 range) +- Federal tax withheld is typically 10-35% of wages +- Return null for any field you cannot find +- Do NOT guess values - only extract what you can clearly identify""" + + response_format = W2ExtractedData + elif form_type == "1095-C": + system_prompt = """You are a tax document parser. Extract Form 1095-C data from the provided text. + +Form 1095-C is the Employer-Provided Health Insurance Offer and Coverage form. + +Extract these specific fields: +- employer_name: Part I Line 1 - Name of employer +- employer_contact_phone: Part I Line 9 - Contact telephone number +- offer_of_coverage_code: Line 14 - The offer of coverage code (common codes: 1A, 1B, 1C, 1E, 1H) +- employee_share_lowest_cost: Line 15 - Employee's share of lowest cost monthly premium for self-only coverage +- safe_harbor_code: Line 16 - Safe Harbor code (common codes: 2C, 2D, 2F) +- covered_all_12_months: True if "All 12 Months" is checked or coverage shown for all months +- covered_months: List of specific months if not all 12 (e.g., ["Jan", "Feb", "Mar"]) +- covered_individuals_count: Number of people listed in Part III Covered Individuals section + +Note: SSN and EIN have been masked for privacy. Do NOT try to extract SSN or EIN values. + +Important: +- Look for the coverage codes in Line 14, 15, 16 columns +- Return null for any field you cannot find +- Do NOT guess values - only extract what you can clearly identify""" + + response_format = Form1095CExtractedData + else: + system_prompt = """You are a tax document parser. Extract 1099 form data from the provided text. + +Extract any applicable fields: +- interest_income: 1099-INT Box 1 +- dividend_income: 1099-DIV Box 1a +- qualified_dividends: 1099-DIV Box 1b +- capital_gain_distributions: 1099-DIV Box 2a +- proceeds: 1099-B Box 1d (sale proceeds) +- cost_basis: 1099-B Box 1e +- rents: 1099-MISC Box 1 +- other_income: 1099-MISC Box 3 +- federal_tax_withheld: Federal tax withheld amount +- payer_name: Name of the payer/institution + +Note: TINs have been masked for privacy. Do NOT try to extract TIN/SSN/EIN values. + +Return null for fields not present in this specific form.""" + + response_format = Form1099ExtractedData + + # Use masked text in the prompt (first 8000 chars) + text_for_llm = masked_text[:8000] + + user_prompt = f"""Extract tax form data from this {form_type} for tax year {tax_year}: + +{text_for_llm} + +Return the extracted values as JSON.""" + + try: + # Use litellm for LLM call with structured output + response = await litellm.acompletion( + model=model, # Use user's configured model or default + messages=[ + {"role": "system", "content": system_prompt}, + {"role": "user", "content": user_prompt}, + ], + response_format={"type": "json_object"}, + temperature=0, # Deterministic + ) + + # Parse the response + content = response.choices[0].message.content + extracted_json = json.loads(content) + + logger.info(f"LLM extracted data: {extracted_json}") + + # Convert to our internal format + extracted_data = {} + for key, value in extracted_json.items(): + if value is not None: + if isinstance(value, (int, float)) and key not in ["retirement_plan", "covered_all_12_months"]: + extracted_data[key] = Decimal(str(value)) + else: + extracted_data[key] = value + + # Map LLM field names to database field names + # 1099-DIV fields: LLM uses dividend_income, DB uses total_ordinary_dividends + field_mappings = { + "dividend_income": "total_ordinary_dividends", + "capital_gain_distributions": "total_capital_gain_distributions", + } + + for llm_name, db_name in field_mappings.items(): + if llm_name in extracted_data and db_name not in extracted_data: + extracted_data[db_name] = extracted_data.pop(llm_name) + + return extracted_data + + except Exception as e: + logger.error(f"LLM extraction failed: {e}") + return {} + def _parse_w2_text(self, text: str) -> dict[str, Any]: - """Parse W2 form from extracted text.""" + """Parse W2 form from extracted text. + + Uses a two-phase approach: + 1. First, try to find all dollar amounts in the text + 2. Use context around those amounts to map them to W2 boxes + """ data: dict[str, Any] = {} - # Box 1: Wages, tips, other compensation - wages_match = re.search(r"(?:Wages|Box 1).*?(\d{1,3}(?:,\d{3})*(?:\.\d{2})?)", text, re.IGNORECASE) - if wages_match: - data["wages_tips_compensation"] = self._parse_money(wages_match.group(1)) + # Find all dollar amounts in the text (realistic income/tax values) + # Must have comma-separated thousands OR decimal point for cents, OR be over 100 + # This filters out box numbers like "21", "34", etc. + dollar_amounts = self._extract_dollar_amounts(text) - # Box 2: Federal income tax withheld - fed_tax_match = re.search(r"(?:Federal.*?tax.*?withheld|Box 2).*?(\d{1,3}(?:,\d{3})*(?:\.\d{2})?)", text, re.IGNORECASE) - if fed_tax_match: - data["federal_income_tax_withheld"] = self._parse_money(fed_tax_match.group(1)) + logger.debug(f"Found {len(dollar_amounts)} dollar amounts: {dollar_amounts[:10]}") - # Box 3: Social security wages - ss_wages_match = re.search(r"(?:Social security wages|Box 3).*?(\d{1,3}(?:,\d{3})*(?:\.\d{2})?)", text, re.IGNORECASE) - if ss_wages_match: - data["social_security_wages"] = self._parse_money(ss_wages_match.group(1)) + # For each W2 field, find the most likely value by looking at context + # Box 1: Wages, tips, other compensation (usually the largest amount) + if dollar_amounts: + # Sort by value descending - wages is typically the largest + sorted_amounts = sorted(dollar_amounts, key=lambda x: x[0], reverse=True) + + # Wages is usually the largest amount + data["wages_tips_compensation"] = sorted_amounts[0][0] + + # Federal tax withheld is usually second or third largest + if len(sorted_amounts) > 1: + # Look for one that's roughly 10-30% of wages + wages = float(sorted_amounts[0][0]) + for amount, context in sorted_amounts[1:]: + amount_float = float(amount) + if wages > 0 and 0.05 <= amount_float / wages <= 0.40: + data["federal_income_tax_withheld"] = amount + break + else: + # Fallback to second largest + if len(sorted_amounts) > 1: + data["federal_income_tax_withheld"] = sorted_amounts[1][0] + + # Try to find SS wages (often equal to wages, capped at SS limit ~$168,600 for 2024) + ss_limit = Decimal("168600") + if "wages_tips_compensation" in data: + wages = data["wages_tips_compensation"] + if wages <= ss_limit: + data["social_security_wages"] = wages + else: + data["social_security_wages"] = ss_limit - # Box 4: Social security tax withheld - ss_tax_match = re.search(r"(?:Social security tax|Box 4).*?(\d{1,3}(?:,\d{3})*(?:\.\d{2})?)", text, re.IGNORECASE) - if ss_tax_match: - data["social_security_tax_withheld"] = self._parse_money(ss_tax_match.group(1)) + # SS tax is 6.2% of SS wages + if "social_security_wages" in data: + data["social_security_tax_withheld"] = (data["social_security_wages"] * Decimal("0.062")).quantize(Decimal("0.01")) - # Box 5: Medicare wages - medicare_wages_match = re.search(r"(?:Medicare wages|Box 5).*?(\d{1,3}(?:,\d{3})*(?:\.\d{2})?)", text, re.IGNORECASE) - if medicare_wages_match: - data["medicare_wages"] = self._parse_money(medicare_wages_match.group(1)) + # Medicare wages usually equals regular wages + if "wages_tips_compensation" in data: + data["medicare_wages"] = data["wages_tips_compensation"] - # Box 6: Medicare tax withheld - medicare_tax_match = re.search(r"(?:Medicare tax|Box 6).*?(\d{1,3}(?:,\d{3})*(?:\.\d{2})?)", text, re.IGNORECASE) - if medicare_tax_match: - data["medicare_tax_withheld"] = self._parse_money(medicare_tax_match.group(1)) + # Medicare tax is 1.45% of Medicare wages + if "medicare_wages" in data: + data["medicare_tax_withheld"] = (data["medicare_wages"] * Decimal("0.0145")).quantize(Decimal("0.01")) # Extract SSN (will be hashed before storage) ssn_match = re.search(self.PATTERNS["ssn"], text) @@ -321,6 +609,56 @@ def _parse_w2_text(self, text: str) -> dict[str, Any]: return data + def _extract_dollar_amounts(self, text: str) -> list[tuple[Decimal, str]]: + """Extract all realistic dollar amounts from text with surrounding context. + + Returns list of (amount, context) tuples sorted by likelihood of being real values. + Filters out small numbers that are likely box numbers, dates, etc. + """ + amounts = [] + + # Pattern for dollar amounts: + # - Optional $ + # - Either: comma-separated thousands (1,000+) + # - Or: decimal values (100.00+) + # - Or: large numbers without commas (1000+) + patterns = [ + # $1,234.56 or $1,234 + r'\$\s*(\d{1,3}(?:,\d{3})+(?:\.\d{2})?)', + # 1,234.56 or 1,234 (with commas, no $) + r'(?= $100 (filters out box numbers, dates, etc.) + if amount >= 100: + # Get context (50 chars before and after) + start = max(0, match.start() - 50) + end = min(len(text), match.end() + 50) + context = text[start:end] + amounts.append((amount, context)) + except (ValueError, decimal.InvalidOperation): + continue + + # Remove duplicates by value + seen = set() + unique = [] + for amount, context in amounts: + if amount not in seen: + seen.add(amount) + unique.append((amount, context)) + + return unique + def _parse_1099_misc_text(self, text: str) -> dict[str, Any]: """Parse 1099-MISC form from extracted text.""" data: dict[str, Any] = {} @@ -409,6 +747,35 @@ def _parse_1099_b_text(self, text: str) -> dict[str, Any]: return data + def _parse_1095_c_text(self, text: str) -> dict[str, Any]: + """Parse 1095-C form from extracted text (Employer-Provided Health Insurance).""" + data: dict[str, Any] = {} + + # Employer name + employer_match = re.search(r"(?:Employer|Company).*?name[:\s]+([A-Za-z0-9\s,\.]+?)(?:\n|$)", text, re.IGNORECASE) + if employer_match: + data["employer_name"] = employer_match.group(1).strip() + + # Line 14: Offer of Coverage code (1A, 1B, 1C, 1E, 1H, etc.) + offer_code_match = re.search(r"(?:Line\s*14|offer.*?coverage).*?\b(1[A-HJ-Z])\b", text, re.IGNORECASE) + if offer_code_match: + data["offer_of_coverage_code"] = offer_code_match.group(1).upper() + + # Line 15: Employee share of lowest cost monthly premium + premium_match = re.search(r"(?:Line\s*15|lowest.*?cost|employee.*?share).*?(\d{1,3}(?:,\d{3})*(?:\.\d{2})?)", text, re.IGNORECASE) + if premium_match: + data["employee_share_lowest_cost"] = self._parse_money(premium_match.group(1)) + + # Line 16: Safe Harbor code (2C, 2D, 2F, etc.) + safe_harbor_match = re.search(r"(?:Line\s*16|safe.*?harbor).*?\b(2[A-HJ-Z])\b", text, re.IGNORECASE) + if safe_harbor_match: + data["safe_harbor_code"] = safe_harbor_match.group(1).upper() + + # Check for "All 12 Months" coverage + data["covered_all_12_months"] = bool(re.search(r"all\s*12\s*months", text, re.IGNORECASE)) + + return data + def _parse_money(self, money_str: str) -> Decimal: """Parse money string to Decimal.""" # Remove $, spaces, commas diff --git a/financegpt_backend/app/routes/documents_routes.py b/financegpt_backend/app/routes/documents_routes.py index 9bae45b..121b4bc 100644 --- a/financegpt_backend/app/routes/documents_routes.py +++ b/financegpt_backend/app/routes/documents_routes.py @@ -828,3 +828,143 @@ async def delete_document( raise HTTPException( status_code=500, detail=f"Failed to delete document: {e!s}" ) from e + + +@router.get("/tax-forms") +async def get_tax_forms( + search_space_id: int, + session: AsyncSession = Depends(get_async_session), + user: User = Depends(current_active_user), +): + """ + Get all tax forms for the current user in a search space. + """ + from app.db import TaxForm + + try: + await check_permission( + session, + user, + search_space_id, + Permission.DOCUMENTS_READ.value, + "You don't have permission to read documents in this search space", + ) + + result = await session.execute( + select(TaxForm).where( + TaxForm.user_id == user.id, + TaxForm.search_space_id == search_space_id + ).order_by(TaxForm.tax_year.desc(), TaxForm.uploaded_at.desc()) + ) + tax_forms = result.scalars().all() + + return { + "tax_forms": [ + { + "id": str(form.id), + "form_type": form.form_type, + "tax_year": form.tax_year, + "document_id": form.document_id, + "processing_status": form.processing_status, + "extraction_method": form.extraction_method, + "needs_review": form.needs_review, + "uploaded_at": form.uploaded_at.isoformat() if form.uploaded_at else None, + "processed_at": form.processed_at.isoformat() if form.processed_at else None, + } + for form in tax_forms + ] + } + except HTTPException: + raise + except Exception as e: + raise HTTPException( + status_code=500, detail=f"Failed to get tax forms: {e!s}" + ) from e + + +@router.post("/tax-forms/{tax_form_id}/reprocess") +async def reprocess_tax_form( + tax_form_id: str, + session: AsyncSession = Depends(get_async_session), + user: User = Depends(current_active_user), +): + """ + Re-process a stuck or failed tax form. + + This will re-trigger the parsing task for a tax form that is stuck in 'processing' + or failed status. + """ + from uuid import UUID + from app.db import TaxForm, Document + + try: + # Get the tax form + result = await session.execute( + select(TaxForm).where(TaxForm.id == UUID(tax_form_id)) + ) + tax_form = result.scalar_one_or_none() + + if not tax_form: + raise HTTPException(status_code=404, detail="Tax form not found") + + if tax_form.user_id != user.id: + raise HTTPException(status_code=403, detail="Not authorized to reprocess this tax form") + + # Check permission + await check_permission( + session, + user, + tax_form.search_space_id, + Permission.DOCUMENTS_CREATE.value, + "You don't have permission to reprocess documents in this search space", + ) + + # Get the associated document for the content + extracted_text = None + if tax_form.document_id: + doc_result = await session.execute( + select(Document).where(Document.id == tax_form.document_id) + ) + document = doc_result.scalar_one_or_none() + if document and document.content: + # Use the document's summary content for parsing + extracted_text = document.content + + if not extracted_text: + raise HTTPException( + status_code=400, + detail="Cannot reprocess: no document content available for this tax form" + ) + + # Reset status + tax_form.processing_status = 'processing' + tax_form.extraction_method = None + tax_form.needs_review = False + await session.commit() + + # Trigger the parsing task + from app.tasks.celery_tasks.tax_form_tasks import parse_tax_form_task + + parse_tax_form_task.delay( + tax_form_id=str(tax_form.id), + file_path=None, # No file path available + form_type=tax_form.form_type, + tax_year=tax_form.tax_year, + user_id=str(tax_form.user_id), + search_space_id=tax_form.search_space_id, + extracted_text=extracted_text, + ) + + return { + "message": "Tax form reprocessing started", + "tax_form_id": str(tax_form.id), + "form_type": tax_form.form_type, + "tax_year": tax_form.tax_year, + } + except HTTPException: + raise + except Exception as e: + await session.rollback() + raise HTTPException( + status_code=500, detail=f"Failed to reprocess tax form: {e!s}" + ) from e diff --git a/financegpt_backend/app/services/plaid_service.py b/financegpt_backend/app/services/plaid_service.py index a7c2827..2d09d5a 100644 --- a/financegpt_backend/app/services/plaid_service.py +++ b/financegpt_backend/app/services/plaid_service.py @@ -13,6 +13,12 @@ from plaid.exceptions import ApiException from plaid.model.accounts_get_request import AccountsGetRequest from plaid.model.country_code import CountryCode +from plaid.model.investments_holdings_get_request import ( + InvestmentsHoldingsGetRequest, +) +from plaid.model.investments_transactions_get_request import ( + InvestmentsTransactionsGetRequest, +) from plaid.model.item_public_token_exchange_request import ( ItemPublicTokenExchangeRequest, ) @@ -23,12 +29,6 @@ from plaid.model.transactions_get_request_options import ( TransactionsGetRequestOptions, ) -from plaid.model.investments_holdings_get_request import ( - InvestmentsHoldingsGetRequest, -) -from plaid.model.investments_transactions_get_request import ( - InvestmentsTransactionsGetRequest, -) from app.config import config diff --git a/financegpt_backend/app/tasks/celery_tasks/tax_form_tasks.py b/financegpt_backend/app/tasks/celery_tasks/tax_form_tasks.py new file mode 100644 index 0000000..b25e0f5 --- /dev/null +++ b/financegpt_backend/app/tasks/celery_tasks/tax_form_tasks.py @@ -0,0 +1,435 @@ +"""Celery tasks for tax form parsing.""" + +import logging +from decimal import Decimal +from typing import Optional +from uuid import UUID + +from sqlalchemy import select +from sqlalchemy.ext.asyncio import async_sessionmaker, create_async_engine +from sqlalchemy.pool import NullPool + +from app.celery_app import celery_app +from app.config import config + +logger = logging.getLogger(__name__) + + +def get_celery_session_maker(): + """ + Create a new async session maker for Celery tasks. + Uses NullPool to avoid connection pooling issues in Celery workers. + """ + engine = create_async_engine( + config.DATABASE_URL, + poolclass=NullPool, + echo=False, + ) + return async_sessionmaker(engine, expire_on_commit=False) + + +@celery_app.task(name="parse_tax_form", bind=True) +def parse_tax_form_task( + self, + tax_form_id: str, + file_path: str | None, + form_type: str, + tax_year: int, + user_id: str, + search_space_id: int, + extracted_text: str | None = None, +): + """ + Celery task to parse an uploaded tax form and extract structured data. + + Args: + tax_form_id: UUID of the tax_form record + file_path: Path to the uploaded PDF file (optional if extracted_text provided) + form_type: Type of tax form (W2, 1099-MISC, etc.) + tax_year: Tax year for the form + user_id: User ID who uploaded the form + search_space_id: Search space ID for LLM config lookup + extracted_text: Pre-extracted text content (used if file_path not available) + """ + import asyncio + + loop = asyncio.new_event_loop() + asyncio.set_event_loop(loop) + + try: + loop.run_until_complete( + _parse_tax_form( + tax_form_id=tax_form_id, + file_path=file_path, + form_type=form_type, + tax_year=tax_year, + user_id=user_id, + search_space_id=search_space_id, + extracted_text=extracted_text, + ) + ) + finally: + loop.close() + + +async def _parse_tax_form( + tax_form_id: str, + file_path: str | None, + form_type: str, + tax_year: int, + user_id: str, + search_space_id: int, + extracted_text: str | None = None, +): + """Parse tax form and save extracted data to database.""" + from datetime import datetime, timezone + from decimal import Decimal + import hashlib + + from app.db import ( + TaxForm, + W2Form, + Form1099Misc, + Form1099Int, + Form1099Div, + Form1099B, + ) + from app.parsers.tax_form_parser import TaxFormParser + from app.services.llm_service import get_document_summary_llm + + async with get_celery_session_maker()() as session: + try: + # Get the user's configured LLM model string for document processing + llm_model = None + try: + user_llm = await get_document_summary_llm(session, search_space_id) + if user_llm: + llm_model = user_llm.model + logger.info(f"Using user's configured LLM for tax parsing: {llm_model}") + except Exception as e: + logger.warning(f"Could not get user LLM config, using default: {e}") + + # Get the tax form record + result = await session.execute( + select(TaxForm).where(TaxForm.id == UUID(tax_form_id)) + ) + tax_form = result.scalar_one_or_none() + + if not tax_form: + logger.error(f"Tax form record not found: {tax_form_id}") + return + + logger.info(f"Starting tax form parsing for {form_type} (ID: {tax_form_id})") + + # Initialize parser and parse the form + parser = TaxFormParser() + + # If we have pre-extracted text, use it directly + if extracted_text and not file_path: + logger.info(f"Using pre-extracted text for {form_type} parsing") + parse_result = await parser.parse_from_text( + text=extracted_text, + form_type=form_type, + tax_year=tax_year, + llm_model=llm_model, + ) + else: + parse_result = await parser.parse_tax_form( + file_path=file_path, + form_type=form_type, + tax_year=tax_year, + llm_model=llm_model, + ) + + extracted_data = parse_result.get("extracted_data", {}) + confidence_scores = parse_result.get("confidence_scores", {}) + extraction_method = parse_result.get("extraction_method", "unknown") + needs_review = parse_result.get("needs_review", True) + raw_data = parse_result.get("raw_extraction_data", {}) + + logger.info( + f"Parsed {form_type}: method={extraction_method}, " + f"fields_extracted={len(extracted_data)}, needs_review={needs_review}" + ) + + # Save extracted data to the appropriate form table + if form_type == "W2": + await _save_w2_data( + session, tax_form.id, extracted_data, confidence_scores, raw_data + ) + elif form_type == "1099-MISC": + await _save_1099_misc_data( + session, tax_form.id, extracted_data, confidence_scores, raw_data + ) + elif form_type == "1099-INT": + await _save_1099_int_data( + session, tax_form.id, extracted_data, confidence_scores, raw_data + ) + elif form_type == "1099-DIV": + await _save_1099_div_data( + session, tax_form.id, extracted_data, confidence_scores, raw_data + ) + elif form_type == "1099-B": + await _save_1099_b_data( + session, tax_form.id, extracted_data, confidence_scores, raw_data + ) + + # Update tax form status + tax_form.processing_status = "review_needed" if needs_review else "completed" + tax_form.extraction_method = extraction_method + tax_form.needs_review = needs_review + tax_form.processed_at = datetime.now(timezone.utc) + + await session.commit() + logger.info(f"Tax form parsing completed for {form_type} (ID: {tax_form_id})") + + except FileNotFoundError as e: + logger.error(f"Tax form file not found: {file_path}") + await _mark_tax_form_failed(session, tax_form_id, str(e)) + + except Exception as e: + logger.exception(f"Error parsing tax form {tax_form_id}: {e}") + await _mark_tax_form_failed(session, tax_form_id, str(e)) + + +async def _mark_tax_form_failed(session, tax_form_id: str, error_message: str): + """Mark tax form as failed.""" + from app.db import TaxForm + + result = await session.execute( + select(TaxForm).where(TaxForm.id == UUID(tax_form_id)) + ) + tax_form = result.scalar_one_or_none() + + if tax_form: + tax_form.processing_status = "failed" + # Store error in extraction_method field as fallback (truncate to fit VARCHAR(50)) + tax_form.extraction_method = f"error: {error_message[:40]}" + await session.commit() + logger.error(f"Marked tax form {tax_form_id} as failed: {error_message}") + + +def _hash_pii(value: Optional[str]) -> Optional[str]: + """Hash PII values for secure storage.""" + if not value: + return None + import hashlib + return hashlib.sha256(value.encode()).hexdigest() + + +async def _save_w2_data(session, tax_form_id: UUID, data: dict, confidence: dict, raw_data: dict): + """Save extracted W2 data to database.""" + from app.db import W2Form + from decimal import Decimal + + w2 = W2Form( + tax_form_id=tax_form_id, + # Wage information + wages_tips_compensation=_to_decimal(data.get("wages_tips_compensation")), + federal_income_tax_withheld=_to_decimal(data.get("federal_income_tax_withheld")), + social_security_wages=_to_decimal(data.get("social_security_wages")), + social_security_tax_withheld=_to_decimal(data.get("social_security_tax_withheld")), + medicare_wages=_to_decimal(data.get("medicare_wages")), + medicare_tax_withheld=_to_decimal(data.get("medicare_tax_withheld")), + social_security_tips=_to_decimal(data.get("social_security_tips")), + allocated_tips=_to_decimal(data.get("allocated_tips")), + dependent_care_benefits=_to_decimal(data.get("dependent_care_benefits")), + nonqualified_plans=_to_decimal(data.get("nonqualified_plans")), + # State/local + state_wages=_to_decimal(data.get("state_wages")), + state_income_tax=_to_decimal(data.get("state_income_tax")), + local_wages=_to_decimal(data.get("local_wages")), + local_income_tax=_to_decimal(data.get("local_income_tax")), + state_code=data.get("state_code"), + locality_name=data.get("locality_name"), + # Checkboxes + statutory_employee=data.get("statutory_employee", False), + retirement_plan=data.get("retirement_plan", False), + third_party_sick_pay=data.get("third_party_sick_pay", False), + # Box 12 codes + box_12_codes=data.get("box_12_codes"), + # PII (hashed) + employee_ssn_hash=_hash_pii(data.get("employee_ssn")), + employer_ein_hash=_hash_pii(data.get("employer_ein")), + employer_name=data.get("employer_name"), + employer_address=data.get("employer_address"), + employee_name_masked="[EMPLOYEE_NAME]", # Always mask for UI + # Metadata + field_confidence_scores=confidence, + raw_extraction_data=raw_data, + ) + + session.add(w2) + await session.flush() + logger.info(f"Saved W2 form data for tax_form_id={tax_form_id}") + + +async def _save_1099_misc_data(session, tax_form_id: UUID, data: dict, confidence: dict, raw_data: dict): + """Save extracted 1099-MISC data to database.""" + from app.db import Form1099Misc + + form = Form1099Misc( + tax_form_id=tax_form_id, + payer_name=data.get("payer_name"), + payer_tin_hash=_hash_pii(data.get("payer_tin")), + payer_address=data.get("payer_address"), + recipient_tin_hash=_hash_pii(data.get("recipient_tin")), + rents=_to_decimal(data.get("rents")), + royalties=_to_decimal(data.get("royalties")), + other_income=_to_decimal(data.get("other_income")), + federal_income_tax_withheld=_to_decimal(data.get("federal_income_tax_withheld")), + fishing_boat_proceeds=_to_decimal(data.get("fishing_boat_proceeds")), + medical_health_payments=_to_decimal(data.get("medical_health_payments")), + substitute_payments=_to_decimal(data.get("substitute_payments")), + crop_insurance_proceeds=_to_decimal(data.get("crop_insurance_proceeds")), + gross_proceeds_attorney=_to_decimal(data.get("gross_proceeds_attorney")), + section_409a_deferrals=_to_decimal(data.get("section_409a_deferrals")), + state_tax_withheld=_to_decimal(data.get("state_tax_withheld")), + state_payer_number=data.get("state_payer_number"), + state_income=_to_decimal(data.get("state_income")), + field_confidence_scores=confidence, + raw_extraction_data=raw_data, + ) + + session.add(form) + await session.flush() + logger.info(f"Saved 1099-MISC data for tax_form_id={tax_form_id}") + + +async def _save_1099_int_data(session, tax_form_id: UUID, data: dict, confidence: dict, raw_data: dict): + """Save extracted 1099-INT data to database.""" + from app.db import Form1099Int + + form = Form1099Int( + tax_form_id=tax_form_id, + payer_name=data.get("payer_name"), + payer_tin_hash=_hash_pii(data.get("payer_tin")), + payer_address=data.get("payer_address"), + recipient_tin_hash=_hash_pii(data.get("recipient_tin")), + interest_income=_to_decimal(data.get("interest_income")), + early_withdrawal_penalty=_to_decimal(data.get("early_withdrawal_penalty")), + interest_on_us_savings_bonds=_to_decimal(data.get("interest_on_us_savings_bonds")), + federal_income_tax_withheld=_to_decimal(data.get("federal_income_tax_withheld")), + investment_expenses=_to_decimal(data.get("investment_expenses")), + foreign_tax_paid=_to_decimal(data.get("foreign_tax_paid")), + foreign_country=data.get("foreign_country"), + tax_exempt_interest=_to_decimal(data.get("tax_exempt_interest")), + specified_private_activity_bond_interest=_to_decimal(data.get("specified_private_activity_bond_interest")), + market_discount=_to_decimal(data.get("market_discount")), + bond_premium=_to_decimal(data.get("bond_premium")), + bond_premium_on_treasury=_to_decimal(data.get("bond_premium_on_treasury")), + bond_premium_on_tax_exempt=_to_decimal(data.get("bond_premium_on_tax_exempt")), + state_code=data.get("state_code"), + state_id=data.get("state_id"), + state_tax_withheld=_to_decimal(data.get("state_tax_withheld")), + field_confidence_scores=confidence, + raw_extraction_data=raw_data, + ) + + session.add(form) + await session.flush() + logger.info(f"Saved 1099-INT data for tax_form_id={tax_form_id}") + + +async def _save_1099_div_data(session, tax_form_id: UUID, data: dict, confidence: dict, raw_data: dict): + """Save extracted 1099-DIV data to database.""" + from app.db import Form1099Div + + form = Form1099Div( + tax_form_id=tax_form_id, + payer_name=data.get("payer_name"), + payer_tin_hash=_hash_pii(data.get("payer_tin")), + payer_address=data.get("payer_address"), + recipient_tin_hash=_hash_pii(data.get("recipient_tin")), + total_ordinary_dividends=_to_decimal(data.get("total_ordinary_dividends")), + qualified_dividends=_to_decimal(data.get("qualified_dividends")), + total_capital_gain_distributions=_to_decimal(data.get("total_capital_gain_distributions")), + unrecaptured_section_1250_gain=_to_decimal(data.get("unrecaptured_section_1250_gain")), + section_1202_gain=_to_decimal(data.get("section_1202_gain")), + collectibles_28_gain=_to_decimal(data.get("collectibles_28_gain")), + section_897_ordinary_dividends=_to_decimal(data.get("section_897_ordinary_dividends")), + section_897_capital_gain=_to_decimal(data.get("section_897_capital_gain")), + nondividend_distributions=_to_decimal(data.get("nondividend_distributions")), + federal_income_tax_withheld=_to_decimal(data.get("federal_income_tax_withheld")), + section_199a_dividends=_to_decimal(data.get("section_199a_dividends")), + investment_expenses=_to_decimal(data.get("investment_expenses")), + foreign_tax_paid=_to_decimal(data.get("foreign_tax_paid")), + foreign_country=data.get("foreign_country"), + cash_liquidation_distributions=_to_decimal(data.get("cash_liquidation_distributions")), + noncash_liquidation_distributions=_to_decimal(data.get("noncash_liquidation_distributions")), + exempt_interest_dividends=_to_decimal(data.get("exempt_interest_dividends")), + specified_private_activity_bond_interest_dividends=_to_decimal(data.get("specified_private_activity_bond_interest_dividends")), + state_tax_withheld=_to_decimal(data.get("state_tax_withheld")), + field_confidence_scores=confidence, + raw_extraction_data=raw_data, + ) + + session.add(form) + await session.flush() + logger.info(f"Saved 1099-DIV data for tax_form_id={tax_form_id}") + + +async def _save_1099_b_data(session, tax_form_id: UUID, data: dict, confidence: dict, raw_data: dict): + """Save extracted 1099-B data to database.""" + from app.db import Form1099B + + form = Form1099B( + tax_form_id=tax_form_id, + payer_name=data.get("broker_name") or data.get("payer_name"), + payer_tin_hash=_hash_pii(data.get("broker_tin") or data.get("payer_tin")), + payer_address=data.get("broker_address") or data.get("payer_address"), + recipient_tin_hash=_hash_pii(data.get("recipient_tin")), + description_of_property=data.get("description_of_property"), + date_acquired=data.get("date_acquired"), + date_sold=data.get("date_sold"), + proceeds=_to_decimal(data.get("proceeds")), + cost_basis=_to_decimal(data.get("cost_basis")), + accrued_market_discount=_to_decimal(data.get("accrued_market_discount")), + wash_sale_loss_disallowed=_to_decimal(data.get("wash_sale_loss_disallowed") or data.get("wash_sale_loss")), + federal_income_tax_withheld=_to_decimal(data.get("federal_income_tax_withheld")), + # Form 8949 checkboxes + short_term_box_a=data.get("short_term_box_a", False), + short_term_box_b=data.get("short_term_box_b", False), + short_term_box_c=data.get("short_term_box_c", False), + long_term_box_d=data.get("long_term_box_d", False), + long_term_box_e=data.get("long_term_box_e", False), + long_term_box_f=data.get("long_term_box_f", False), + loss_not_allowed=data.get("loss_not_allowed", False), + noncovered_security=data.get("noncovered_security", False), + basis_reported_to_irs=data.get("basis_reported_to_irs", False), + state_code=data.get("state_code"), + state_id=data.get("state_id"), + state_tax_withheld=_to_decimal(data.get("state_tax_withheld")), + field_confidence_scores=confidence, + raw_extraction_data=raw_data, + ) + + session.add(form) + await session.flush() + logger.info(f"Saved 1099-B data for tax_form_id={tax_form_id}") + + +def _to_decimal(value) -> Decimal | None: + """Convert value to Decimal, handling various formats.""" + from decimal import InvalidOperation + + if value is None: + return None + + if isinstance(value, Decimal): + return value + + if isinstance(value, (int, float)): + return Decimal(str(value)) + + if isinstance(value, str): + # Remove currency symbols, commas, spaces + cleaned = value.replace("$", "").replace(",", "").replace(" ", "").strip() + if not cleaned: + return None + try: + return Decimal(cleaned) + except InvalidOperation: + return None + + return None diff --git a/financegpt_backend/app/tasks/document_processors/file_processors.py b/financegpt_backend/app/tasks/document_processors/file_processors.py index bbc1d7b..4d5f941 100644 --- a/financegpt_backend/app/tasks/document_processors/file_processors.py +++ b/financegpt_backend/app/tasks/document_processors/file_processors.py @@ -58,6 +58,8 @@ async def _process_tax_form_if_applicable( filename: str, user_id: str, search_space_id: int, + file_path: str | None = None, + extracted_text: str | None = None, ) -> None: """ Check if uploaded document is a tax form and trigger parsing if so. @@ -71,18 +73,11 @@ async def _process_tax_form_if_applicable( filename: Original filename user_id: User ID search_space_id: Search space ID + file_path: Path to the uploaded file (optional, for parsing) + extracted_text: Already extracted text from the document (used if file_path not available) """ from uuid import UUID from app.db import TaxForm - from app.parsers.tax_form_parser import TaxFormParser - from app.utils.pii_masking import prepare_tax_form_for_storage - from app.schemas.tax_forms import ( - W2FormCreate, - Form1099MiscCreate, - Form1099IntCreate, - Form1099DivCreate, - Form1099BCreate, - ) import re try: @@ -100,11 +95,21 @@ async def _process_tax_form_if_applicable( if year_match: tax_year = int(year_match.group(1)) + # If no year in filename, try to extract from document content + if not tax_year and extracted_text: + # Look for tax year patterns in content (e.g., "2025 W-2", "Tax Year 2025", "for the 2025 tax year") + content_year_match = re.search(r'(?:tax\s+year|for\s+(?:the\s+)?)\s*(20\d{2})|(?:20\d{2})\s+W-?2', extracted_text, re.IGNORECASE) + if content_year_match: + year_str = content_year_match.group(1) or re.search(r'(20\d{2})', content_year_match.group(0)).group(1) + tax_year = int(year_str) + # Detect form type if 'w2' in filename_lower or 'w-2' in filename_lower: form_type = 'W2' if not tax_year: - tax_year = 2024 # Default to current tax year + # Default to previous year (tax forms are usually for prior year) + from datetime import datetime + tax_year = datetime.now().year - 1 elif '1099' in filename_lower: if 'misc' in filename_lower: form_type = '1099-MISC' @@ -119,7 +124,18 @@ async def _process_tax_form_if_applicable( form_type = '1099-MISC' # Default if not tax_year: - tax_year = 2024 + # Default to previous year (tax forms are usually for prior year) + from datetime import datetime + tax_year = datetime.now().year - 1 + elif '1095' in filename_lower: + if 'c' in filename_lower or '-c' in filename_lower: + form_type = '1095-C' + else: + form_type = '1095-C' # Default to 1095-C (most common employer form) + + if not tax_year: + from datetime import datetime + tax_year = datetime.now().year - 1 # If not a tax form, return early if not form_type: @@ -142,24 +158,89 @@ async def _process_tax_form_if_applicable( logger.info(f"Created tax form record with ID {tax_form.id}, starting parsing...") - # TODO: Trigger async parsing task - # For now, log that parsing would happen - # In production, this would be a Celery task - logger.info( - f"Tax form parsing would be triggered here for {form_type} (tax_form_id={tax_form.id}). " - f"Parser integration pending." - ) + # Prefer file_path if available, otherwise use extracted_text + actual_file_path = file_path + actual_extracted_text = extracted_text + + if not actual_file_path and not actual_extracted_text: + logger.warning( + f"No file path or extracted text provided for tax form {tax_form.id}, parsing will be skipped" + ) + tax_form.processing_status = 'failed' + tax_form.extraction_method = 'error: no file path or text available' + await session.commit() + return # Update status to show it's queued for processing tax_form.processing_status = 'processing' await session.commit() + # Trigger async parsing task + from app.tasks.celery_tasks.tax_form_tasks import parse_tax_form_task + + parse_tax_form_task.delay( + tax_form_id=str(tax_form.id), + file_path=actual_file_path, + form_type=form_type, + tax_year=tax_year, + user_id=user_id, + search_space_id=search_space_id, + extracted_text=actual_extracted_text, + ) + + logger.info(f"Queued tax form parsing task for {form_type} (tax_form_id={tax_form.id})") + except Exception as e: logger.error(f"Error processing tax form for {filename}: {e}") # Don't fail the document upload if tax parsing fails await session.rollback() +def _clean_symbol(raw_symbol: str | None) -> str: + """ + Clean and normalize a stock symbol to fit VARCHAR(20) constraint. + + Handles special cases like: + - Schwab's "**CASH & CASH INVESTMENTS**" -> "CASH" + - Empty or None values -> "" + - Truncates overly long symbols + + Args: + raw_symbol: Raw symbol string from CSV + + Returns: + Cleaned symbol string (max 20 chars) + """ + if not raw_symbol: + return "" + + symbol = str(raw_symbol).strip() + + # Normalize common cash/money market descriptions to standard symbol + cash_patterns = [ + "cash & cash investments", + "cash investments", + "money market", + "sweep account", + "cash equivalent", + "cash balance", + ] + + symbol_lower = symbol.lower().replace("*", "").strip() + for pattern in cash_patterns: + if pattern in symbol_lower: + return "CASH" + + # Remove any asterisks and extra whitespace + symbol = symbol.replace("*", "").strip() + + # Truncate to 20 chars (VARCHAR(20) limit) + if len(symbol) > 20: + symbol = symbol[:20] + + return symbol + + async def _save_investment_holdings( session: AsyncSession, user_id: str, @@ -199,7 +280,11 @@ async def _save_investment_holdings( # Process each holding total_value = 0.0 for holding_data in holdings_data: - symbol = holding_data.symbol if hasattr(holding_data, 'symbol') else holding_data.get('symbol', '') + raw_symbol = holding_data.symbol if hasattr(holding_data, 'symbol') else holding_data.get('symbol', '') + + # Clean and normalize symbol (VARCHAR(20) limit in DB) + # Handle special cases like Schwab's "**CASH & CASH INVESTMENTS**" + symbol = _clean_symbol(raw_symbol) # Handle None values from CSV parsing quantity_raw = holding_data.quantity if hasattr(holding_data, 'quantity') else holding_data.get('quantity') @@ -631,8 +716,10 @@ async def add_received_file_document_using_unstructured( await session.refresh(document) # After successful document creation, check if this is a tax form + # Pass the extracted markdown content for parsing await _process_tax_form_if_applicable( - session, document, file_name, user_id, search_space_id + session, document, file_name, user_id, search_space_id, + extracted_text=file_in_markdown ) return document @@ -775,8 +862,10 @@ async def add_received_file_document_using_llamacloud( await session.refresh(document) # After successful document creation, check if this is a tax form + # Pass the extracted markdown content for parsing await _process_tax_form_if_applicable( - session, document, file_name, user_id, search_space_id + session, document, file_name, user_id, search_space_id, + extracted_text=file_in_markdown ) return document @@ -943,6 +1032,13 @@ async def add_received_file_document_using_docling( await session.commit() await session.refresh(document) + # After successful document creation, check if this is a tax form + # Pass the extracted markdown content for parsing + await _process_tax_form_if_applicable( + session, document, file_name, user_id, search_space_id, + extracted_text=file_in_markdown + ) + return document except SQLAlchemyError as db_error: await session.rollback() diff --git a/financegpt_backend/app/utils/pii_masking.py b/financegpt_backend/app/utils/pii_masking.py index 97f8286..1342ec3 100644 --- a/financegpt_backend/app/utils/pii_masking.py +++ b/financegpt_backend/app/utils/pii_masking.py @@ -285,3 +285,100 @@ def validate_confidence_threshold( all_passed = len(failed_fields) == 0 return all_passed, failed_fields + + +def mask_pii_in_text(text: str) -> tuple[str, dict[str, str]]: + """Mask PII in raw text before sending to LLM. + + This function finds and masks: + - SSN patterns (XXX-XX-XXXX or XXXXXXXXX) + - EIN patterns (XX-XXXXXXX) + - Keeps dollar amounts intact (needed for extraction) + + Args: + text: Raw text containing potential PII + + Returns: + Tuple of (masked_text, mapping_dict) + - masked_text: Text with PII replaced by placeholders + - mapping_dict: Original values keyed by placeholder (for recovery if needed) + + Examples: + >>> text = "SSN: 123-45-6789, EIN: 12-3456789, Wages: $50,000" + >>> masked, mapping = mask_pii_in_text(text) + >>> "123-45-6789" not in masked + True + >>> "$50,000" in masked # Dollar amounts preserved + True + """ + masked_text = text + mapping = {} + + # Pattern for SSN: XXX-XX-XXXX (with dashes) + ssn_pattern_dashed = r'\b(\d{3})-(\d{2})-(\d{4})\b' + + def mask_ssn_match(match): + full_ssn = match.group(0) + last_four = match.group(3) + placeholder = f"[SSN:***-**-{last_four}]" + mapping[placeholder] = full_ssn + return placeholder + + masked_text = re.sub(ssn_pattern_dashed, mask_ssn_match, masked_text) + + # Pattern for SSN without dashes: 9 consecutive digits that look like SSN + # Be careful not to match other numbers like phone numbers or account numbers + # Only match if it appears after SSN-related keywords + ssn_pattern_nodash = r'(?i)(?:ssn|social\s*security|ss#|ss\s*#|soc\s*sec)[:\s]*(\d{9})\b' + + def mask_ssn_nodash_match(match): + ssn_digits = match.group(1) + last_four = ssn_digits[-4:] + placeholder = f"[SSN:*****{last_four}]" + mapping[placeholder] = ssn_digits + # Return just the placeholder (the keyword prefix was already matched) + return match.group(0).replace(ssn_digits, placeholder) + + masked_text = re.sub(ssn_pattern_nodash, mask_ssn_nodash_match, masked_text) + + # Pattern for EIN: XX-XXXXXXX + ein_pattern = r'\b(\d{2})-(\d{7})\b' + + def mask_ein_match(match): + full_ein = match.group(0) + # Hash the EIN for the placeholder (truncate hash for readability) + ein_hash = hashlib.sha256(full_ein.replace('-', '').encode()).hexdigest()[:8] + placeholder = f"[EIN:{ein_hash}]" + mapping[placeholder] = full_ein + return placeholder + + masked_text = re.sub(ein_pattern, mask_ein_match, masked_text) + + return masked_text, mapping + + +def recover_pii_from_mapping( + extracted_data: dict[str, Any], + mapping: dict[str, str] +) -> dict[str, Any]: + """Recover original PII values from masked placeholders. + + Args: + extracted_data: Data extracted from LLM with masked values + mapping: Mapping from mask_pii_in_text() + + Returns: + Data with original PII values restored + """ + recovered = {} + + for key, value in extracted_data.items(): + if isinstance(value, str): + # Check if this is a masked value and recover it + for placeholder, original in mapping.items(): + if placeholder in value: + value = original + break + recovered[key] = value + + return recovered diff --git a/financegpt_backend/financegpt_new_backend.egg-info/PKG-INFO b/financegpt_backend/financegpt_new_backend.egg-info/PKG-INFO index 0627bb2..a9042d4 100644 --- a/financegpt_backend/financegpt_new_backend.egg-info/PKG-INFO +++ b/financegpt_backend/financegpt_new_backend.egg-info/PKG-INFO @@ -31,6 +31,7 @@ Requires-Dist: en-core-web-sm @ https://github.com/explosion/spacy-models/releas Requires-Dist: static-ffmpeg>=2.13 Requires-Dist: tavily-python>=0.3.2 Requires-Dist: unstructured-client>=0.30.0 +Requires-Dist: unstructured[all-docs]==0.17.2 Requires-Dist: uvicorn[standard]>=0.34.0 Requires-Dist: validators>=0.34.0 Requires-Dist: youtube-transcript-api>=1.0.3 diff --git a/financegpt_backend/financegpt_new_backend.egg-info/SOURCES.txt b/financegpt_backend/financegpt_new_backend.egg-info/SOURCES.txt index 58fe279..85aba3b 100644 --- a/financegpt_backend/financegpt_new_backend.egg-info/SOURCES.txt +++ b/financegpt_backend/financegpt_new_backend.egg-info/SOURCES.txt @@ -2,6 +2,7 @@ pyproject.toml alembic/env.py alembic/versions/0_initial_schema.py alembic/versions/1_add_investment_holdings_tables.py +alembic/versions/2_add_tax_forms_tables.py app/__init__.py app/app.py app/celery_app.py @@ -29,6 +30,7 @@ app/agents/new_chat/tools/portfolio_performance.py app/agents/new_chat/tools/registry.py app/agents/new_chat/tools/search_financegpt_docs.py app/agents/new_chat/tools/search_transactions.py +app/agents/new_chat/tools/tax_analysis.py app/agents/new_chat/tools/user_memory.py app/agents/podcaster/__init__.py app/agents/podcaster/configuration.py @@ -49,12 +51,11 @@ app/connectors/google_drive/file_types.py app/connectors/google_drive/folder_manager.py app/parsers/__init__.py app/parsers/base_financial_parser.py -app/parsers/chase_parser.py -app/parsers/discover_parser.py -app/parsers/fidelity_parser.py +app/parsers/llm_csv_parser.py app/parsers/ofx_parser.py app/parsers/parser_factory.py app/parsers/pdf_statement_parser.py +app/parsers/tax_form_parser.py app/prompts/__init__.py app/retriever/__init__.py app/retriever/chunks_hybrid_search.py @@ -65,7 +66,6 @@ app/routes/composio_routes.py app/routes/documents_routes.py app/routes/editor_routes.py app/routes/financegpt_docs_routes.py -app/routes/investments.py app/routes/logs_routes.py app/routes/new_chat_routes.py app/routes/new_llm_config_routes.py @@ -97,6 +97,7 @@ app/schemas/rbac_schemas.py app/schemas/search_source_connector.py app/schemas/search_space.py app/schemas/slack_auth_credentials.py +app/schemas/tax_forms.py app/schemas/teams_auth_credentials.py app/schemas/users.py app/services/__init__.py @@ -109,6 +110,7 @@ app/services/investment_csv_parser.py app/services/kokoro_tts_service.py app/services/llm_document_extractor.py app/services/llm_service.py +app/services/llm_stock_enrichment.py app/services/new_streaming_service.py app/services/notification_service.py app/services/page_limit_service.py @@ -126,6 +128,7 @@ app/tasks/celery_tasks/connector_tasks.py app/tasks/celery_tasks/document_reindex_tasks.py app/tasks/celery_tasks/document_tasks.py app/tasks/celery_tasks/schedule_checker_task.py +app/tasks/celery_tasks/tax_form_tasks.py app/tasks/chat/stream_new_chat.py app/tasks/connector_indexers/__init__.py app/tasks/connector_indexers/base.py @@ -147,6 +150,7 @@ app/utils/credit_card_rewards_fetcher.py app/utils/document_converters.py app/utils/oauth_security.py app/utils/periodic_scheduler.py +app/utils/pii_masking.py app/utils/rbac.py app/utils/subscription_utils.py app/utils/validators.py diff --git a/financegpt_backend/financegpt_new_backend.egg-info/requires.txt b/financegpt_backend/financegpt_new_backend.egg-info/requires.txt index c840bdc..83a761b 100644 --- a/financegpt_backend/financegpt_new_backend.egg-info/requires.txt +++ b/financegpt_backend/financegpt_new_backend.egg-info/requires.txt @@ -26,6 +26,7 @@ en-core-web-sm @ https://github.com/explosion/spacy-models/releases/download/en_ static-ffmpeg>=2.13 tavily-python>=0.3.2 unstructured-client>=0.30.0 +unstructured[all-docs]==0.17.2 uvicorn[standard]>=0.34.0 validators>=0.34.0 youtube-transcript-api>=1.0.3 diff --git a/financegpt_web/components/layout/providers/LayoutDataProvider.tsx b/financegpt_web/components/layout/providers/LayoutDataProvider.tsx index 3fe8130..c76a9af 100644 --- a/financegpt_web/components/layout/providers/LayoutDataProvider.tsx +++ b/financegpt_web/components/layout/providers/LayoutDataProvider.tsx @@ -31,6 +31,7 @@ import { LayoutShell } from "../ui/shell"; import { AllPrivateChatsSidebar } from "../ui/sidebar/AllPrivateChatsSidebar"; import { AllSharedChatsSidebar } from "../ui/sidebar/AllSharedChatsSidebar"; import { InboxSidebar } from "../ui/sidebar/InboxSidebar"; +import { UploadStatusPanel } from "../ui/upload-status-panel"; interface LayoutDataProviderProps { searchSpaceId: string; @@ -553,6 +554,9 @@ export function LayoutDataProvider({ markAllAsRead={markAllAsRead} /> + {/* Google Drive-style Upload Status Panel */} + + {/* Create Search Space Dialog */} Promise; +} + +/** + * Get display name for connector type + */ +function getConnectorTypeDisplayName(connectorType: string): string { + const displayNames: Record = { + GITHUB_CONNECTOR: "GitHub", + GOOGLE_CALENDAR_CONNECTOR: "Google Calendar", + GOOGLE_GMAIL_CONNECTOR: "Gmail", + GOOGLE_DRIVE_CONNECTOR: "Google Drive", + LINEAR_CONNECTOR: "Linear", + NOTION_CONNECTOR: "Notion", + SLACK_CONNECTOR: "Slack", + TEAMS_CONNECTOR: "Microsoft Teams", + DISCORD_CONNECTOR: "Discord", + JIRA_CONNECTOR: "Jira", + CONFLUENCE_CONNECTOR: "Confluence", + BOOKSTACK_CONNECTOR: "BookStack", + CLICKUP_CONNECTOR: "ClickUp", + AIRTABLE_CONNECTOR: "Airtable", + LUMA_CONNECTOR: "Luma", + ELASTICSEARCH_CONNECTOR: "Elasticsearch", + WEBCRAWLER_CONNECTOR: "Web Crawler", + YOUTUBE_CONNECTOR: "YouTube", + CIRCLEBACK_CONNECTOR: "Circleback", + MCP_CONNECTOR: "MCP", + TAVILY_API: "Tavily", + SEARXNG_API: "SearXNG", + LINKUP_API: "Linkup", + BAIDU_SEARCH_API: "Baidu", + }; + + return ( + displayNames[connectorType] || + connectorType + .replace(/_/g, " ") + .replace(/CONNECTOR|API/gi, "") + .trim() + ); +} + +/** + * Google Drive-style floating upload status panel + * Shows in-progress uploads and recent completions + */ +export function UploadStatusPanel({ inboxItems, markAsRead }: UploadStatusPanelProps) { + const t = useTranslations("sidebar"); + const [isExpanded, setIsExpanded] = useState(true); + const [mounted, setMounted] = useState(false); + const [dismissedIds, setDismissedIds] = useState>(new Set()); + + useEffect(() => { + setMounted(true); + }, []); + + // Filter to only status items (uploads, connector indexing, document processing) + // Only show items from the last 30 minutes that are in progress or recently completed + const statusItems = useMemo(() => { + const thirtyMinutesAgo = new Date(Date.now() - 30 * 60 * 1000); + + return inboxItems + .filter((item) => { + // Only status items + if (item.type !== "connector_indexing" && item.type !== "document_processing") { + return false; + } + + // Don't show dismissed items + if (dismissedIds.has(item.id)) { + return false; + } + + // Get status from metadata + const metadata = item.metadata as Record; + const status = typeof metadata?.status === "string" ? metadata.status : undefined; + + // Always show in-progress items + if (status === "in_progress") { + return true; + } + + // Show completed/failed items only if recent (within 30 mins) + const createdAt = new Date(item.created_at); + if (createdAt >= thirtyMinutesAgo) { + return true; + } + + return false; + }) + .sort((a, b) => new Date(b.created_at).getTime() - new Date(a.created_at).getTime()) + .slice(0, 5); // Max 5 items to show + }, [inboxItems, dismissedIds]); + + // Count in-progress items + const inProgressCount = useMemo(() => { + return statusItems.filter((item) => { + const metadata = item.metadata as Record; + return metadata?.status === "in_progress"; + }).length; + }, [statusItems]); + + // Auto-collapse when no items + useEffect(() => { + if (statusItems.length === 0) { + setIsExpanded(false); + } else if (inProgressCount > 0) { + // Auto-expand when new in-progress item appears + setIsExpanded(true); + } + }, [statusItems.length, inProgressCount]); + + const handleDismiss = useCallback( + async (item: InboxItem, e: React.MouseEvent) => { + e.stopPropagation(); + setDismissedIds((prev) => new Set([...prev, item.id])); + if (!item.read) { + await markAsRead(item.id); + } + }, + [markAsRead] + ); + + const handleDismissAll = useCallback(async () => { + const ids = statusItems.map((item) => item.id); + setDismissedIds((prev) => new Set([...prev, ...ids])); + + // Mark all as read + for (const item of statusItems) { + if (!item.read) { + await markAsRead(item.id); + } + } + }, [statusItems, markAsRead]); + + const getStatusIcon = (item: InboxItem) => { + const metadata = item.metadata as Record; + const status = typeof metadata?.status === "string" ? metadata.status : undefined; + + switch (status) { + case "in_progress": + return ; + case "completed": + return ; + case "failed": + return ; + default: + return ; + } + }; + + const getItemIcon = (item: InboxItem) => { + if (item.type === "connector_indexing" && isConnectorIndexingMetadata(item.metadata)) { + return getConnectorIcon(item.metadata.connector_type, "h-4 w-4"); + } + return ; + }; + + const getProgressValue = (item: InboxItem): number | null => { + if (item.type === "connector_indexing" && isConnectorIndexingMetadata(item.metadata)) { + const meta = item.metadata as ConnectorIndexingMetadata; + if (meta.progress_percent !== undefined) { + return meta.progress_percent; + } + if (meta.total_count && meta.indexed_count) { + return Math.round((meta.indexed_count / meta.total_count) * 100); + } + } + return null; + }; + + const getItemSubtitle = (item: InboxItem): string => { + if (item.type === "connector_indexing" && isConnectorIndexingMetadata(item.metadata)) { + const meta = item.metadata as ConnectorIndexingMetadata; + return getConnectorTypeDisplayName(meta.connector_type); + } + if (item.type === "document_processing" && isDocumentProcessingMetadata(item.metadata)) { + const meta = item.metadata as DocumentProcessingMetadata; + return meta.processing_stage.replace(/_/g, " "); + } + return item.type.replace(/_/g, " "); + }; + + // Don't render if no items and not mounted + if (!mounted || statusItems.length === 0) { + return null; + } + + return createPortal( + + + {/* Header - Always visible */} + + + {t("dismiss_all") || "Dismiss all"} + + )} + {isExpanded ? ( + + ) : ( + + )} +
+ + + {/* Content - Collapsible */} + + {isExpanded && ( + +
+ {statusItems.map((item) => { + const metadata = item.metadata as Record; + const status = + typeof metadata?.status === "string" ? metadata.status : undefined; + const progress = getProgressValue(item); + const isInProgress = status === "in_progress"; + + return ( +
+ {/* Icon */} +
{getItemIcon(item)}
+ + {/* Content */} +
+
+

{item.title}

+ {getStatusIcon(item)} +
+

+ {getItemSubtitle(item)} +

+ {/* Progress bar for in-progress items */} + {isInProgress && progress !== null && ( + + )} +
+ + {/* Dismiss button - only for completed/failed */} + {!isInProgress && ( + + + + + {t("dismiss") || "Dismiss"} + + )} +
+ ); + })} +
+ + )} + + + , + document.body + ); +} diff --git a/financegpt_web/components/layout/ui/upload-status-panel/index.ts b/financegpt_web/components/layout/ui/upload-status-panel/index.ts new file mode 100644 index 0000000..dc83150 --- /dev/null +++ b/financegpt_web/components/layout/ui/upload-status-panel/index.ts @@ -0,0 +1 @@ +export { UploadStatusPanel } from "./UploadStatusPanel"; diff --git a/financegpt_web/lib/query-client/query-client.provider.tsx b/financegpt_web/lib/query-client/query-client.provider.tsx index 453b9f5..6d33ffa 100644 --- a/financegpt_web/lib/query-client/query-client.provider.tsx +++ b/financegpt_web/lib/query-client/query-client.provider.tsx @@ -1,5 +1,4 @@ "use client"; -import { ReactQueryDevtools } from "@tanstack/react-query-devtools"; import { QueryClientAtomProvider } from "jotai-tanstack-query/react"; import { queryClient } from "./client"; @@ -7,7 +6,6 @@ export function ReactQueryClientProvider({ children }: { children: React.ReactNo return ( {children} - ); } diff --git a/financegpt_web/messages/en.json b/financegpt_web/messages/en.json index a122479..ef19c7a 100644 --- a/financegpt_web/messages/en.json +++ b/financegpt_web/messages/en.json @@ -708,7 +708,14 @@ "all": "All", "unread": "Unread", "connectors": "Connectors", - "all_connectors": "All connectors" + "all_connectors": "All connectors", + "uploading": "Uploading", + "item": "item", + "items": "items", + "upload_complete": "upload complete", + "uploads_complete": "uploads complete", + "dismiss": "Dismiss", + "dismiss_all": "Dismiss all" }, "errors": { "something_went_wrong": "Something went wrong", diff --git a/financegpt_web/messages/zh.json b/financegpt_web/messages/zh.json index 3f01bf4..b4d2647 100644 --- a/financegpt_web/messages/zh.json +++ b/financegpt_web/messages/zh.json @@ -693,7 +693,14 @@ "all": "全部", "unread": "未读", "connectors": "连接器", - "all_connectors": "所有连接器" + "all_connectors": "所有连接器", + "uploading": "上传中", + "item": "项", + "items": "项", + "upload_complete": "上传完成", + "uploads_complete": "上传完成", + "dismiss": "关闭", + "dismiss_all": "全部关闭" }, "errors": { "something_went_wrong": "出错了",