AutoCodeRover

Stale
GitHub Python NOASSERTION

Description

AutoCodeRover is a project structure-aware autonomous software engineer agent that achieves automated program repair and issue resolution by understanding the overall codebase architecture.

Key Features

  • Fully autonomous bug fixing and feature addition on GitHub issues using LLM-powered program analysis
  • Program Structure Aware code search via abstract syntax tree analysis for precise context retrieval
  • Statistical fault localization using test cases to achieve higher repair rates when tests are available
  • Supports multiple LLM backends including GPT-4o, Claude, Gemini, Groq, and Anthropic via AWS Bedrock
  • Achieves 46.20% efficacy on SWE-bench Verified with tasks costing less than $0.70 each
  • Three operating modes: GitHub issue, local issue, and SWE-bench benchmark evaluation

Use Cases

💡 Automated resolution of open GitHub issues in open-source software projects
💡 Autonomous program repair for bug fixes in production codebases with test suites
💡 Research benchmarking on SWE-bench for evaluating AI software engineering capabilities
💡 Accelerated feature development through AI-assisted code patch generation

Tags

coding python agent llm automation

Categories

💻 Coding Agent

Quick Start

Set your OpenAI key: export OPENAI_KEY=sk-YOUR-KEY. Build the Docker image: docker build -f Dockerfile.minimal -t acr . Run on a GitHub issue: docker run -it -e OPENAI_KEY acr. Provide the project clone link, commit hash, and issue link. AutoCodeRover will analyze the codebase, locate the fault, and generate a patch automatically.
Visit GitHub

Related Projects