Todozi - AI/Human Task Management System

Todozi is a comprehensive, AI-first task and knowledge management system that enables seamless collaboration between humans and AI agents. Built entirely in Python, Todozi combines powerful task management with semantic understanding, memory systems, and intelligent automation.

🎯 Core Philosophy

Todozi is designed around a TAG-FIRST approach where AI models can naturally embed structured information directly in conversations. This allows models to remember, organize, and manage tasks better than humans - capturing context, dependencies, and relationships automatically.

🚀 Key Features

1. AI-Powered Task Management

• Intelligent task creation with automatic priority inference
• Semantic search across all content types
• AI/human/collaborative task assignment
• Project-based organization with automatic grouping
• Real-time progress tracking and analytics

2. Enhanced Terminal UI (TUI)

• Beautiful, feature-rich terminal interface with 8 comprehensive tabs
• Real-time analytics dashboard with completion tracking
• Modal dialogs for task editing and management
• Toast notifications and activity feed
• Live file watching for automatic updates
• Server management and API key controls

3. Chat Interface for AI Models 🆕

The chat/ directory contains a revolutionary web-based chat interface that demonstrates Todozi's core value proposition:

• 20% Sidebar: Chat history with scrollable session list
• 40% Chat Interface: Real-time messaging with AI assistant
• 40% Todozi Panel: Dynamic task management with project filtering

Why This Matters: The chat interface shows how AI models can use Todozi tags (<todozi>, <memory>, <idea>, etc.) to naturally structure information during conversations. This allows models to:

• Remember context better than humans
• Automatically organize tasks and knowledge
• Maintain project continuity across sessions
• Capture insights and ideas in real-time

4. Training System 🆕

The train/ directory contains powerful training scripts for fine-tuning AI models on Todozi data:

• train-todozi-basic.py: Comprehensive BCO (Best-of-N) training script
• Supports multiple batch strategies (balanced, bootstrap, k-fold, stratified, etc.)
• Integration with Hugging Face models (Gemma, Llama, etc.)
• Automatic dataset loading from batch files
• Configurable training parameters

Training Features:

• Multiple batch file formats and strategies
• BCO (Best-of-N) training with reference model
• Support for bf16/fp16 precision
• Automatic checkpointing and progress tracking
• Integration with TRL library for RLHF

5. Memory & Knowledge System

• Structured memory capture with importance levels
• Short-term and long-term memory retention
• Emotional memory tracking
• Secret/human/standard memory types
• Semantic memory search

6. Idea & Innovation Tracking

• Capture creative ideas with importance levels
• Team/public/private sharing options
• Breakthrough idea identification
• Tag-based organization

7. Advanced Features

• Error logging with severity classification
• Reminder system with overdue detection
• Training data collection for model improvement
• Code chunking for large projects
• Emotional state tracking
• Content summarization

📦 Installation

pip install todozi

Or for development installation:

git clone https://github.com/cyber-boost/todozi.git
cd todozi/tdz_py
pip install -e .

🛠️ Quick Start

Terminal UI (TUI)

Launch the enhanced terminal interface:

todozi
# or
python -m todozi.tui

Keyboard Shortcuts:

• 1-8: Direct tab access (Projects, Tasks, Done, Find, More, API, Feed, Bye)
• a: Add new task
• e: Edit task
• c: Complete task
• d: Delete task
• Space: Task actions menu
• Enter: Task details modal
• Ctrl+E: Modal task editing
• Ctrl+R: Full refresh
• ?: Help

Command Line Interface (CLI)

Use the tdz command for quick operations:

# Launch TUI (default)
tdz

# CLI commands
tdz task "Review quarterly budget"
tdz urgent "Fix critical security issue"
tdz find "authentication"
tdz stats

Python API

from todozi import Storage, Task, Priority, Status

# Initialize storage
storage = await Storage.new()

# Create a task
task = Task.new_full(
    user_id="user123",
    action="Implement user authentication",
    time="4 hours",
    priority=Priority.HIGH,
    parent_project="web-app",
    status=Status.TODO,
    assignee=Assignee.AI,
    tags=["backend", "security"],
    context_notes="Use OAuth2 with JWT tokens"
)

# Add to project
await storage.add_task_to_project(task)

# List tasks
tasks = storage.list_tasks_across_projects(TaskFilters())

# Complete task
storage.complete_task_in_project(task.id)

💬 Chat Interface

The chat interface demonstrates Todozi's tag-based approach for AI models:

cd chat
pip install -r requirements.txt
python chat.py

Then open http://localhost:8275 in your browser.

Example Conversation:

User: "I need to build a user authentication system for my web app"

AI: I'll help you organize this! <todozi>Build user authentication system; 4 hours; high; web-app; todo; ai; backend,security; none; Implement OAuth2 with JWT tokens; 0%</todozi>

<memory>standard; OAuth2 flows are complex; Need secure token handling; Important for web security; high; long; security,authentication</memory>

What specific authentication methods do you want to support?

The AI automatically:

• Creates structured tasks with proper fields
• Captures important memories
• Maintains project context
• Organizes information semantically

🎓 Training AI Models

Train models on Todozi data using the training scripts:

cd train
pip install -r requirements.txt

# Basic training
python pure/train-todozi-basic.py

# Advanced training (if available)
python pure/train-todozi-advanced.py

Training Script Features:

• Automatic batch file selection
• BCO (Best-of-N) training configuration
• Support for multiple model architectures
• Configurable training parameters
• Progress tracking and checkpointing

Available Batch Strategies:

• balanced/: Balanced sampling across categories
• bootstrap/: Bootstrap resampling
• kfold/: K-fold cross-validation splits
• stratified/: Stratified sampling
• progressive/: Progressive training sets
• ml/: Machine learning optimized splits
• custom/: Custom batch configurations

📚 Todozi Tags

Todozi uses XML-like tags that AI models can naturally embed in conversations:

Task Creation

<todozi>action; time; priority; project; status; assignee; tags; dependencies; context; progress%</todozi>

Example:

<todozi>Fix login bug; 2 hours; high; auth-system; in_progress; ai; bug,security; none; JWT validation issue; 25%</todozi>

Memory Capture

<memory>type; moment; meaning; reason; importance; term; tags</memory>

Example:

<memory>standard; Discovered async patterns in Rust; Makes concurrent code cleaner; Will use for API calls; high; long; rust,async,patterns</memory>

Idea Capture

<idea>idea; share; importance; tags; context</idea>

Example:

<idea>Voice-controlled task management; team; breakthrough; innovation,product,ai; Could revolutionize productivity</idea>

Error Logging

<error>title; description; severity; category; source; context; tags</error>

Example:

<error>Database timeout; Connection failed after 30s; critical; database; postgres; Need connection pooling; performance,database</error>

🏗️ Project Structure

tdz_py/
├── todozi/          # Main package
│   ├── tui.py       # Enhanced Terminal UI
│   ├── cli.py       # Command-line interface
│   ├── storage.py   # File-based storage
│   ├── models.py    # Data models
│   ├── server.py    # REST API server
│   └── ...
├── chat/            # Chat interface for AI models
│   ├── chat.py      # Flask chat server
│   ├── system.py    # AI system prompts
│   ├── tdz.html     # Web interface
│   └── requirements.txt
├── train/           # Training scripts
│   ├── pure/        # Pure training scripts
│   │   ├── train-todozi-basic.py
│   │   └── train-todozi-advanced.py
│   ├── batches/     # Training batch files
│   └── requirements.txt
├── tests/           # Test suite
├── tdz.py           # CLI entry point
├── pyproject.toml   # Package configuration
└── py.sh            # Build & publish script

🔧 Development

Building the Package

Use the automated build script:

./py.sh              # Build only
./py.sh --upload     # Build and upload to PyPI

The script automatically:

• Increments version number (patch version)
• Cleans old build artifacts
• Builds source distribution and wheel
• Optionally uploads to PyPI

Running Tests

pytest
# or
python -m pytest

Development Installation

pip install -e ".[dev]"

📊 Features Overview

Terminal UI Tabs

1. 📁 Projects: Manage and organize projects
2. 📋 Tasks: View, filter, and manage tasks with analytics
3. ✅ Done: Completed tasks with sorting and filtering
4. 🔍 Find: Semantic search across all content
5. 🔮 More: Extended data (Ideas, Memories, Feelings, Errors, Training, Queue, Reminders, Analytics)
6. 🔑 API: Server management and API key administration
7. 📰 Feed: Live activity feed with real-time updates
8. 👋 Bye: Exit screen

API Server

Start the REST API server:

from todozi.server import start_server

# Start server on default port
await start_server()

Key Endpoints:

• GET /tasks - List all tasks
• POST /tasks - Create new task
• GET /tasks/search?q={query} - Semantic search
• GET /agents - List AI agents
• POST /chat/process - Process chat with Todozi tags

🎯 Use Cases

For Developers

• Project task management
• Code review tracking
• Bug tracking and resolution
• Documentation tasks
• Learning and knowledge capture

For AI Researchers

• Training data collection
• Model fine-tuning workflows
• Experiment tracking
• Knowledge base building
• Multi-agent coordination

For Teams

• Collaborative task management
• Shared knowledge base
• Project planning and tracking
• Meeting notes and action items
• Innovation capture

🔗 Links

• Website: https://todozi.com
• GitHub: https://github.com/cyber-boost/todozi
• Documentation: https://todozi.com
• Issues: https://github.com/cyber-boost/todozi/issues

📄 License

MIT License - see LICENSE file for details.

🤝 Contributing

Contributions welcome! Please see CONTRIBUTING.md for guidelines.

🙏 Acknowledgments

Built with love for the AI and developer communities. Special thanks to all contributors and users who help make Todozi better every day.

---

Todozi - Where AI and humans collaborate seamlessly. 🚀