prr (PR Resolver)

    /\_____/\
   /  o   o  \
  ( ==  ^  == )
   )         (
  (           )
 ( (  )   (  ) )
(__(__)___(__)__)

CLI tool to automatically resolve LLM review bot comments on PRs using LLM-powered fixing and verification.

Features

• Fetches GitHub Copilot review comments from PRs
• Uses LLM to detect which issues still exist in the code
• Generates fix prompts and runs Cursor CLI or opencode to fix issues
• Verifies fixes with LLM to prevent false positives
• Tracks "lessons learned" to prevent flip-flopping between solutions
• Batched commits to avoid flooding git history
• Hash-based work directories for efficient re-runs

Installation

bun install
bun run build

# Run directly
bun dist/index.js <pr-url>

# Or link globally
bun link
prr --version  # See the cat!

Configuration

Create a .env file (see .env.example):

# Required
GITHUB_TOKEN=ghp_xxxx

# LLM for verification (anthropic or openai)
PRR_LLM_PROVIDER=anthropic
PRR_LLM_MODEL=claude-sonnet-4-20250514
ANTHROPIC_API_KEY=sk-ant-xxxx

# Or use OpenAI
# PRR_LLM_PROVIDER=openai
# PRR_LLM_MODEL=gpt-4o
# OPENAI_API_KEY=sk-xxxx

# Default fixer tool
PRR_TOOL=cursor

# Bot usernames to look for (comma-separated, matched as substrings)
# Defaults to: copilot, coderabbitai, greptile, codex-connector, sourcery, codiumai
# PRR_BOT_USERS=copilot,coderabbitai,greptile

Supported Bots

Out of the box, prr detects comments from:

• GitHub Copilot (Copilot, copilot-pull-request-reviewer[bot])
• CodeRabbit (coderabbitai[bot])
• Greptile (greptile-apps[bot])
• ChatGPT Codex (chatgpt-codex-connector[bot])
• Sourcery (sourcery-ai[bot])
• CodiumAI/Qodo (codiumai[bot])

Add custom bots via PRR_BOT_USERS env var (comma-separated, matched as substrings).

Usage

# Basic usage - fix locally, don't push
prr https://github.com/owner/repo/pull/123

# Shorthand syntax
prr owner/repo#123

# Auto-push mode - full automation loop
prr https://github.com/owner/repo/pull/123 --auto-push

# Use specific fixer tool
prr https://github.com/owner/repo/pull/123 --tool opencode

# Dry run - show issues without fixing
prr https://github.com/owner/repo/pull/123 --dry-run

# Keep work directory for inspection
prr https://github.com/owner/repo/pull/123 --keep-workdir

# All options
prr https://github.com/owner/repo/pull/123 \
  --tool cursor \
  --auto-push \
  --max-fix-iterations 10 \
  --max-push-iterations 3 \
  --poll-interval 120 \
  --keep-workdir \
  --verbose

How It Works

1. Fetch Comments: Gets all LLM review bot comments from the PR via GitHub GraphQL API (supports Copilot, CodeRabbit, Sourcery, etc.)

2. Detect Unresolved: For each comment, asks the verification LLM: "Is this issue still present in the code?" Skips issues already marked as fixed in state.

3. Generate Prompt: Builds a fix prompt including:

   • All unresolved issues with code context
   • "Lessons learned" from previous attempts to prevent flip-flopping

4. Run Fixer: Executes Cursor CLI or opencode with the prompt in the cloned repo

5. Verify Fixes: For each changed file, asks the verification LLM: "Does this diff address the concern?" Updates state accordingly.

6. Commit: Squashes all changes into a single commit with descriptive message

7. Push (optional): In --auto-push mode, pushes and waits for bots to re-review, then loops

Work Directory

• Location: ~/.prr/work/<hash>
• Hash is based on owner/repo#number - same PR reuses same directory
• Cleaned up by default on success
• Use --keep-workdir to preserve for debugging

State File

State is persisted in <workdir>/.pr-resolver-state.json:

{
  "pr": "owner/repo#123",
  "branch": "feature-branch",
  "iterations": [...],
  "lessonsLearned": [
    "Attempted try/catch but issue requires early return",
    "The validateUser function must preserve backward compatibility"
  ],
  "verifiedFixed": ["comment_id_1", "comment_id_2"]
}

Requirements

• Node.js >= 18
• cursor CLI (if using Cursor) or opencode CLI (if using opencode)
• GitHub personal access token with repo scope
• Anthropic or OpenAI API key for verification

License

MIT