OpenHands - chunhualiao/public-docs GitHub Wiki
Overview
Integrated environment with support for
- chatting with agents
- terminal
- coding: str_replace_editor // ?? patching generation and applying the patches.
- web browsing
- Jupyter
Tips
Remember to:
- Keep your tasks small.
- Be as specific as possible.
- Provide as much context as possible.
- Commit and push frequently.
Based on these findings and community feedback, the following models have been verified to work reasonably well with OpenHands:
- claude-3-5-sonnet (recommended)
- gpt-4 / gpt-4o
- llama-3.1-405b // llama-3.3?
- deepseek-v2.5 // deepseek-v3?
Repository configuration
https://docs.all-hands.dev/modules/usage/prompting/customization
You can customize OpenHands' behavior for your repository by creating a .openhands directory in your repository's root. At minimum, it should contain the file .openhands/microagents/repo.md, which includes instructions that will be given to the agent every time it works with this repository.
We suggest including the following information:
- Repository Overview: A brief description of your project's purpose and architecture.
- Directory Structure: Key directories and their purposes.
- Development Guidelines: Project-specific coding standards and practices.
- Testing Requirements: How to run tests and what types of tests are required.
- Setup Instructions: Steps needed to build and run the project.
OpenHands + CodeAct v2.1 (claude-3-5-sonnet-20241022)
OpenHands + CodeAct v2.1 (claude-3-5-sonnet-20241022) is a state-of-the-art open-source software development agent that has achieved remarkable results in solving real-world GitHub issues[1][6]. This latest version, powered by Anthropic's Claude-3.5 model, has demonstrated a 53% resolution rate on SWE-Bench and a 41.7% success rate on SWE-Bench Lite[1][6].
Key Features
-
Advanced Language Model: CodeAct 2.1 utilizes the Claude-3.5 model, significantly improving natural language comprehension and enabling better understanding of developer issues[1].
-
Function Calling: The agent's functionality has been refined to incorporate precise function calling, enhancing task execution accuracy[1].
-
Improved Directory Traversal: Substantial improvements in directory navigation allow the agent to handle larger and more complex problems efficiently[1].
-
Real-World Application: Unlike many AI tools limited to controlled environments, CodeAct 2.1 can autonomously solve actual GitHub issues[1][7].
-
Open-Source Framework: Developers can freely use, improve, and adapt OpenHands, making it a versatile choice for the broader community[7].
Performance
OpenHands + CodeAct v2.1 has set new benchmarks in AI-assisted software engineering:
- Achieved the top position on SWE-Bench Verified, an OpenAI-approved benchmark for coding agents[5][6].
- Outperformed other popular AI coding tools and frameworks[5].
- Demonstrated a 53% resolution rate on SWE-Bench and a 41.7% success rate on SWE-Bench Lite[1][6].
This combination of openness, competitiveness, and real-world applicability has positioned OpenHands + CodeAct v2.1 as a leading choice for developers seeking an effective AI solution for software engineering tasks[7][8].
Citations:
- [1] https://www.reddit.com/r/machinelearningnews/comments/1gh90vw/all_hands_ai_open_sources_openhands_codeact_21_a/
- [2] https://github.com/All-Hands-AI/OpenHands/releases
- [3] https://arxiv.org/html/2412.21139v1
- [4] https://www.youtube.com/watch?v=PJ8XUfivG20
- [5] https://www.youtube.com/watch?v=P7jC0L-hjM0
- [6] https://www.swebench.com
- [7] https://www.youtube.com/watch?v=su067vIv7eQ
- [8] https://nlp.elvissaravia.com/p/ai-agents-weekly-codeact-21-copilot
- [9] https://github.com/All-Hands-AI/OpenHands/issues/4629
- [10] https://www.youtube.com/watch?v=Q3DyeIV96tY
- [11] https://substack.com/@elvissaravia/note/c-75169236
How does the combination (OpenHands + CodeAct+ Claude-3.5) work together ?
The combination of OpenHands, CodeAct 2.1, and Claude-3.5 Sonnet creates a powerful and versatile AI-driven software development platform. Here's how these components work together:
OpenHands Framework
OpenHands serves as the open-source foundation, providing a flexible environment for AI-assisted software engineering[1]. It allows for the deployment of AI agents capable of performing various tasks such as running commands, browsing the web, and interacting with codebases[3].
CodeAct 2.1
CodeAct 2.1 is the latest iteration of the coding-based agent within the OpenHands framework[1]. It enhances the system's capabilities in several ways:
- Task Precision: It utilizes enhanced function calling to improve accuracy in executing complex coding tasks[1].
- Efficiency: CodeAct 2.1 consolidates actions and provides unified action spaces, streamlining the development process[3].
- User Interaction: It improves the interaction between the user and the AI through more sophisticated conversations[3].
Claude-3.5 Sonnet Integration
The integration of Anthropic's Claude-3.5 Sonnet model significantly boosts the platform's performance:
- Advanced Language Understanding: Claude-3.5 Sonnet provides state-of-the-art natural language comprehension, enabling better interpretation of developer instructions[4].
- Coding Prowess: In internal evaluations, Claude-3.5 Sonnet demonstrated the ability to solve 64% of coding problems, showcasing its strong capabilities in software development tasks[4].
- Visual Reasoning: The model excels in visual tasks, which can be particularly useful for interpreting charts, graphs, and other visual elements in software projects[4].
Synergistic Performance
Together, these components create a synergistic effect:
- OpenHands provides the framework and tools for AI-assisted development.
- CodeAct 2.1 offers specialized coding capabilities and efficient task execution.
- Claude-3.5 Sonnet powers the system with advanced language understanding and problem-solving abilities.
This combination has achieved impressive results, including a 53% resolution rate on SWE-Bench and a 41.7% success rate on SWE-Bench Lite, outperforming many other AI coding tools[1].
By leveraging these technologies, developers can tackle complex software engineering tasks with greater efficiency and accuracy, from initial design to bug fixes and optimizations[6].
Citations:
- [1] https://www.youtube.com/watch?v=PJ8XUfivG20
- [2] https://community.sap.com/t5/technology-blogs-by-members/anthropic-s-claude-3-5-sonnet-why-everyone-s-talking-about-it/ba-p/13739653
- [3] https://www.chaindesk.ai/tools/youtube-summarizer/open-hands-best-ai-software-engineer-beats-claude-3-5-sonnet-bolt-new-generate-full-stack-apps-PJ8XUfivG20
- [4] https://www.anthropic.com/news/claude-3-5-sonnet
- [5] https://docs.github.com/en/copilot/using-github-copilot/using-claude-sonnet-in-github-copilot
- [6] https://www.anthropic.com/claude/sonnet
- [7] https://www.geeky-gadgets.com/openhands-ai-platform-software-engineering/