Local Automation Agent which transforms natural language into executable system tasks with AI-powered planning and code generation.

TaskWeaver is a command line automation agent that bridges the gap between human intent and system execution. Simply describe what you want to do, and TaskWeaver will:

1. Plan : Generate a clear step-by-step execution strategy
2. Code : Create safe, runnable Python code
3. Execute : Run the code with your approval and safety controls

Perfect for automating repetitive tasks, file organization, data processing, and system maintenance—all from your terminal.

• Natural Language Interface —> Describe tasks in plain English
• Local LLM Integration —> Uses Ollama for on-device AI (privacy-first)
• Safety First Execution:
  • Code validation (blocks dangerous patterns)
  • User confirmation before running
  • Execution timeout protection
  • Secure temporary file handling
• Rich Terminal UI —> Beautiful, readable output with syntax highlighting
• Customizable Settings —> Choose your model and timeout limits

• Python 3.8 or higher
• Ollama installed and running (download here: https://ollama.ai)
• A model available locally (e.g., ollama pull llama3)

git clone https://github.com/yourusername/taskweaver.git
cd taskweaver
pip install -r [requirements.txt]

python main.py "organize my downloads by file type"

python main.py "create a backup of my documents" --model llama3

python main.py "process all csv files" --timeout 60

$ python main.py "count all python files in my projects"

─ TaskWeaver ─────────────────────
Local Automation Agent
───────────────────────────────────

Task: count all python files in my projects

─ Execution Plan ──────────────────
1. Search for .py files recursively
2. Count total files found
3. Display results by directory
───────────────────────────────────

─ Generated Code ──────────────────
import os
from pathlib import Path

count = 0
for root, dirs, files in os.walk(Path.home()):
    for file in files:
        if file.endswith('.py'):
            count += 1
print(f"Total Python files: {count}")
───────────────────────────────────

Execute this code? [y/n]: y

Executing...

Found 1,247 Python files across 12 projects

typer==0.9.0
rich==13.7.0
ollama==0.1.0

TaskWeaver implements multiple safety layers:

Code Pattern Validation Blocks:

• os.system()
• subprocess.Popen(), subprocess.call()
• eval(), exec()
• import() User Confirmation —> All generated code must be approved before execution

Execution Timeout —> Prevents infinite loops (default: 30 seconds)

Secure Temp Files —> Uses tempfile.NamedTemporaryFile() with automatic cleanup

taskweaver/
├── main.py              # Core agent orchestration
├── requirements.txt     # Python dependencies
├── README.md            # Documentation
├── test_sample.txt
└── assets/
|   └── TaskWeaver.png
|   └── diagram.svg

File Organization —> Sort, rename, and organize files

Data Processing —> Parse, transform, and aggregate data

Bulk Operations —> Batch rename, convert, or compress files

Directory Cleanup —> Find duplicates, old files, large folders

System Maintenance —> Generate reports, backups, health checks

• Requires a local Ollama instance running
• Code generation quality depends on the LLM model used
• Some complex tasks may need manual refinement
• Not suitable for live system-critical operations without thorough testing

Contributions are welcome. Please follow these steps:

• Fork the repository
• Create a feature branch (git checkout -b feature/your-idea)
• Commit your changes (git commit -m "Add your feature")
• Push to the branch (git push origin feature/your-idea)
• Open a Pull Request

• Ollama Documentation: https://github.com/jmorganca/ollama
• Typer CLI Framework: https://typer.tiangolo.com/
• Rich Terminal Library: https://rich.readthedocs.io/

• Multi-step task chaining
• Web UI dashboard
• Task history and logging
• Custom code templates
• Integration with external APIs
• Execution dry-run mode

• Report bugs on GitHub Issues
• Share ideas and ask questions in Discussions