LedgerLlama 🦙 — The Post-SaaS Dashboard

A fully local-first, zero-backend, SQLite-WASM powered finance and habit intelligence dashboard with local AI analysis.

LedgerLlama runs entirely within your browser. It combines a highly-stylized modern UI built with React + Vite + Tailwind CSS with the sheer power of an embedded SQLite WASM database via OPFS. In addition to personal finance and habit tracking, it uses @mlc-ai/web-llm to process data directly in the browser to offer LLM-powered categorization and insights!

✨ Features

• Dashboard & UI Framework

  • Clean, modern, responsive glassmorphism UI using Tailwind CSS (var(--background), var(--card), etc.).
  • Integrated lucide-react for beautiful iconography.
  • Dynamic routing to sub-pages: Dashboard, Transactions, Habits, Intelligence, and Settings.

• SQLite WASM & OPFS

  • Database resides in an Origin Private File System (sqlite3_opfs_async_vfs), allowing persistent fast reads/writes without ever needing a backend.
  • Full schema setup (schema.sql) for transactions, categories, habits, and logs.

• Transaction Management

  • Implemented standard CRUD (Create, Read, Update, Delete) UI for financial transactions.
  • Securely log incomes and expenses manually.
  • Smart Search: Integrated fuzzy filtering into the Transactions grid to easily sort by category and description.

• Habit Tracker system

  • Track daily consistencies on a 7-day trailing axis.
  • Clean UI that maps boolean progress to habits inside the local SQLite database.

• Charting Insights Layer

  • Visualize 30-day expense trends with Chart.js.
  • Doughnut charts representing categorical breakdowns.

• CSV Import Utility

  • PapaParse integration allows seamless drag-and-drop or file uploading of banking CSVs.

• WebLLM AI Intelligence

  • Embedded @mlc-ai/web-llm running inside a Web Worker.
  • Downloads an 8-B parameter instruction-tuned model (e.g., Llama-3-8B) to your GPU cache.
  • Complete data privacy: 100% of LLM queries happen locally on your own hardware!
  • AI Budget Planner: Embedded a full conversational chat inside the Intelligence view using WebLLM for immediate financial advice.
  • Batch categorizes uncategorized transactions and offers an "AI Financial Advisor" for 30-day analyses.

• Data Export, Privacy & Security

  • Export the raw .sqlite3 OPFS DB via the File System Access API.
  • CSV and JSON backups allow users to download raw .csv or .json reports.
  • Security Module: Includes a LockScreen wrapper monitoring OS-level idle properties (mousemove, keydown) to lock the app after 5 minutes of inactivity with a PIN.
  • AES-GCM Encryption: Optional Web Crypto API encryption when exporting backups.

• PWA Enablement & Performance

  • vite-plugin-pwa caching for CSS/JS bundles.
  • React lazy loading out-of-the-box (Suspense) for splitting Intelligence models into chunks.
  • SQL indexes targeting frequently requested aggregations.

🚀 Getting Started

To run LedgerLlama locally:

# Install dependencies
npm install

# Run the development server
npm run dev

Visit http://localhost:5173 to view the application.

📦 Build for Production

npm run build
npm run preview

🛠 Tech Stack

• Frontend Core: React, TypeScript, Vite
• Styling: Tailwind CSS, PostCSS, Lucide React
• Database: SQLite WASM, OPFS (Origin Private File System)
• AI/ML: @mlc-ai/web-llm
• Charting: Chart.js, react-chartjs-2
• Data parsing: PapaParse

📄 License

This project is licensed under the MIT License - see the LICENSE file for details.