Awesome AI AgentsCode Automation and Testing

OpenAutoCoder/Agentless

⭐ 2109 Python repository created 2024-06-30

Agentless is an innovative open-source project designed to automatically solve software development problems without relying on traditional agent-based methods. It employs a structured three-phase process to address issues in software code: localization, repair, and patch validation. The localization phase hierarchically narrows down faults from files to specific classes or functions, and finally to precise edit locations. In the repair phase, Agentless generates multiple candidate patches in a simple diff format for each identified bug. The patch validation phase involves selecting relevant regression tests and generating additional reproduction tests to confirm the original error, then re-ranking the candidate patches based on test outcomes to select the best fix for submission. Agentless has demonstrated strong performance on the SWE-bench lite benchmark, achieving a 27.3% fix rate with an average cost of $0.34 per issue, and has been integrated with Claude 3.5 Sonnet to further improve solve rates. The project is actively maintained, with recent releases enhancing its capabilities and performance. It provides comprehensive setup instructions, including environment creation and API key configuration, enabling users to reproduce experiments and run the tool on SWE-bench problems. The repository also offers access to complete artifacts from its runs on SWE-bench benchmarks and manual classifications of problem sets. Agentless is well-documented, with a detailed README that includes news updates, comparison graphs against other approaches, citation information, and acknowledgments to related projects. This project is valuable for researchers and practitioners interested in automated software debugging and repair, leveraging large language models and hierarchical fault localization techniques to improve software reliability efficiently and cost-effectively.

https://github.com/OpenAutoCoder/Agentless

agentagentlessartificial-intelligenceautomated-software-repairautomatic-bug-fixingcandidate-patchesclaude-3.5-sonnetfault-localizationhierarchical-localizationllmopen-sourcepatch-validationregression-testingsoftware-debuggingsoftware-developmentsoftware-development-problemssoftware-engineering-agentssoftware-repairswe-bench

Also in Code Automation and Testing

alibaba/arthas

Arthas is an open-source Java diagnostic tool by Alibaba that enables real-time production troubleshooting and performance monitoring without code changes or JVM restarts.

SWE-agent/SWE-agent

SWE-agent is an autonomous software engineering tool that uses large language models to automatically fix GitHub issues, perform web tasks, and solve cybersecurity challenges, achieving state-of-the-art results in automated software repair and offensive cybersecurity.

qodo-ai/qodo-cover

Qodo-Cover is an AI-powered tool that automates the generation of unit tests to enhance code coverage, supporting integration with GitHub CI workflows and local CLI usage.

judge0/judge0

Judge0 is a robust, scalable, and open-source online code execution system that supports over 60 programming languages and is designed for easy integration into various applications.

BuilderIO/micro-agent

BuilderIO/micro-agent is an AI-powered tool that writes and iteratively refines code by generating tests and ensuring all test cases pass, focusing on reliable test-driven development.

mattzcarey/shippie

Shippie is an AI-powered extensible code review agent that integrates into CI/CD pipelines to automatically detect code issues and improve software delivery speed.

the-open-engine/zeroshot

Command-line orchestrator that drives coding agent CLIs through an executor-verifier loop, so the agent that wrote a change is never the one that declares it correct.

AnandChowdhary/continuous-claude

Shell CLI that runs Claude Code in a continuous loop, opening a pull request per iteration, watching CI, merging or discarding the work, and carrying context forward through a shared markdown file.