I Built an AI That Turns GitHub Issues Into Pull Requests — No Local Setup Required

Short summary

Resolvo is a multi-agent pipeline built on LangGraph that takes a GitHub issue and repo URL, then returns a working pull request with passing tests and adversarial review. It routes work by confidence level, uses Gemini Flash for critical reasoning steps and lite models for cheaper tasks, grounds planning with live web search, and runs tests in E2B sandboxes. The author claims 85% reduction in issue turnaround time for lightweight bugs.

• •Multi-agent pipeline (LangGraph + Gemini) converts GitHub issues to PRs with tests
• •Confidence-based routing and tiered model usage optimize cost vs reasoning depth
• •Real sandbox execution and RRF-based retrieval ground the agent's output

Generated with AI, which can make mistakes.