DiaryML

Private, encrypted AI journaling with emotion detection and deep insights. Zero surveillance, zero tracking, 100% local.

Desktop/Web Interface	Mobile App

Full-featured web interface	Mobile companion app with offline sync

✨ Features

🔒 AES-256 Encrypted - SQLCipher database, zero cloud dependencies
🧠 Local AI - Any GGUF model (1B-3B optimized), CPU-only, hotswappable
😊 Emotion Detection - Calibrated AI analyzes 6 emotions with conversational awareness
📊 Deep Analytics - Writing streaks, productivity scores, temporal patterns
💬 Chat Sessions - Multiple conversations with AI, full history
🔍 Advanced Search - Full-text, date range, emotion filters, RAG semantic search
✏️ Entry Editing - Edit past entries with emotion re-analysis
💾 Backup/Restore - One-click zip backup
🔊 Voice Output - Browser TTS for AI responses
📱 Mobile App - Flutter app with offline sync (see mobile app)

🚀 Quick Start

Note: pysqlcipher3 is required for encryption. Without it, your database will NOT be encrypted!

Option 1: Docker (Recommended - Easiest!)

# 1. Download a GGUF model to models/ directory
# 2. Build and run with Docker Compose
docker-compose up -d

# 3. Open http://localhost:8000 and set your password

Docker Benefits:

✅ No dependency hell - everything bundled
✅ Consistent environment across platforms
✅ Easy updates with docker-compose pull && docker-compose up -d
✅ Data persists in volumes (diary.db, models/, chroma_db/)

Option 2: Manual Installation

Install: pip install -r requirements.txt
Get a Model: Download any GGUF model (1-3B recommended) to models/
Run: Windows: start.bat | Linux/macOS: ./start.sh
Open: Navigate to http://localhost:8000 and set your password

🧠 Supported Models

Works with any GGUF model - optimized for 1-3B on CPU:

1B (fastest) - 2k tokens, 24k context
2B (balanced) - 3k tokens, 28k context
3B (quality) - 4k tokens, 32k context
Auto-detects thinking models (<think> tags) and vision models

🔧 Tech Stack

FastAPI + SQLCipher + llama.cpp + ChromaDB + Hugging Face transformers

🔐 Why DiaryML?

Surveillance-Free & Private:

✅ Zero tracking - No telemetry, no analytics, no phone-home
✅ Zero cloud - Everything runs locally on your machine
✅ Zero accounts - No sign-ups, no services, no third parties
✅ AES-256 encrypted - SQLCipher database secured with your password
✅ Open source - Audit the code, know exactly what it does

Easy to Use:

✅ One command with Docker (docker-compose up -d)
✅ Or simple scripts (start.bat / start.sh)
✅ Works on any platform (Windows, Linux, macOS)
✅ No GPU required - runs on CPU only

Easy to Modify:

✅ Simple Python codebase (FastAPI backend)
✅ Vanilla JavaScript frontend (no complex frameworks)
✅ Well-documented API endpoints
✅ MIT licensed - modify freely

⚠️ Important: Requires pysqlcipher3 for encryption. Without it, data is unencrypted!

📦 Requirements

Python 3.10+ | 2-4GB RAM | ~1-3GB disk | No GPU required
Keyboard shortcuts: Ctrl+F (search), Ctrl+S (save), Ctrl+L (lock), Esc (close)

🐛 Troubleshooting

Model not loading? Check models/ folder for valid GGUF file
Slow responses? Use smaller 1B model or Q4_K_M quantization
Can't unlock? First password creates DB; delete diary.db to reset
ChromaDB error? Delete chroma_db/ folder, auto-rebuilds on restart

📄 License

MIT License - use freely, modify, share.

🙏 Credits

Built with FastAPI, ChromaDB, llama.cpp, sentence-transformers, and Hugging Face transformers.

DiaryML: Your private creative companion. Capture emotions that words cannot express.

4.4 KiB Raw Permalink Blame History