dothething  by fluffypony

Autonomous AI agent for local task execution

Created 1 month ago
1,646 stars

Top 25.1% on SourcePulse

GitHubView on GitHub
Project Summary

Summary

fluffypony/dothething provides a local, autonomous AI agent designed to execute complex tasks described in plain English. It targets users needing to automate research, data extraction, web automation, file manipulation, and code execution, offering a powerful, extensible solution for complex workflows.

How It Works

The agent routes Claude Opus through OpenRouter, enabling it to intelligently select and utilize a suite of tools. Core components include a local SearXNG instance for web searches and Notte/Camoufox for advanced browser automation, capable of handling captchas and multi-step interactions. It supports file operations, shell commands, and HTTP requests. Novelty lies in its extensibility via custom skills loaded from ~/.dtt/skills/ and integration with MCP servers. Cost management features like prompt caching and delegation to cheaper models (Sonnet) optimize resource usage, while thread persistence allows resuming interrupted tasks.

Quick Start & Requirements

Installation involves cloning the repository and running the ./dtt.sh script. Key requirements include macOS or Linux, Python 3.11+, and an OpenRouter API key. Optional dependencies for enhanced functionality include 2Captcha and AgentMail API keys. Linux clipboard support necessitates wl-clipboard or xclip. The initial run requires several minutes for environment setup (Python venv, SearXNG, Notte), with subsequent startups being rapid. The project website is dotheth.ing.

Highlighted Details

  • Autonomous execution of multi-step tasks based on natural language prompts.
  • Robust local web research (SearXNG) and interactive browser automation (Notte/Camoufox).
  • Extensible architecture supporting custom skills and MCP server integrations.
  • Features for cost control, including prompt caching and model delegation.
  • Orchestrator mode enables parallel execution and management of multiple agents.
  • Accepts mid-task user input and can prompt for clarification.

Maintenance & Community

The README does not detail specific contributors, sponsorships, or community channels like Discord or Slack, beyond the primary GitHub repository.

Licensing & Compatibility

The project is licensed under the permissive BSD 3-Clause license, generally allowing for commercial use and integration into closed-source projects without significant restrictions.

Limitations & Caveats

Core functionality is dependent on external API keys (OpenRouter), with optional keys required for advanced features. Linux clipboard integration requires specific package installations. Some user interface elements, such as the TUI, are noted as experimental.

Health Check
Last Commit

2 weeks ago

Responsiveness

Inactive

Pull Requests (30d)
0
Issues (30d)
0
Star History
1,677 stars in the last 30 days

Explore Similar Projects

Starred by Yiran Wu Yiran Wu(Coauthor of AutoGen), Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), and
3 more.

OS-Copilot by OS-Copilot

0.1%
2k
OS agent for automating daily tasks
Created 2 years ago
Updated 1 year ago
Starred by Vincent Weisser Vincent Weisser(Cofounder of Prime Intellect) and Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems").

ii-agent by Intelligent-Internet

0.1%
3k
Agentic framework for building and deploying intelligent agents
Created 1 year ago
Updated 1 month ago
Starred by Chip Huyen Chip Huyen(Author of "AI Engineering", "Designing Machine Learning Systems"), Abubakar Abid Abubakar Abid(Cofounder of Gradio), and
3 more.

owl by camel-ai

0.1%
20k
Multi-agent framework for real-world task automation
Created 1 year ago
Updated 4 days ago
Feedback? Help us improve.