Revu - AI-Powered Code Review Assistant

Revu is a GitHub App that leverages LLMs to provide intelligent, context-aware code reviews for pull requests. By analyzing the entire codebase and changes, Revu offers comprehensive feedback that goes beyond simple style checks.

Features

• Extended Thinking - Enhanced reasoning for deeper code analysis and security review
• Contextual Analysis - Understands changes in the context of the entire codebase
• Precise Code Suggestions - Suggests accurate code changes and improvements
• PR Validation - Automatically skips problematic PRs with helpful feedback
• Customizable - Configurable coding guidelines, branch filters, and file exclusions

Quick Start

Installation

# Use the correct Node.js version
nvm use v23.7.0

# Install dependencies
yarn install

GitHub App Setup

1. Create a GitHub App at Settings > Developer settings > GitHub Apps

2. Configure permissions and events:

   Webhook URL: Your server URL or smee.io proxy
   Permissions:
     - Pull requests: Read & write
     - Contents: Read
     - Organization members: Read # required for org membership checks
   Events:
     - Pull request
     - Pull request review
     - Pull request review comment # required for threaded discussion replies

3. Save your App ID, Private Key, and Webhook Secret

Proxy User Setup

Since GitHub Apps cannot receive review requests directly, Revu uses a proxy user:

1. Create a dedicated GitHub user account (e.g., revu-bot-reviewer)
2. Generate a personal access token with repository access
3. Ensure the proxy user has read access to target repositories

Environment Configuration

Create a .env file with the following variables:

# Required
ANTHROPIC_API_KEY=your_anthropic_key
APP_ID=your_github_app_id
PRIVATE_KEY_PATH=path/to/private-key.pem
WEBHOOK_SECRET=your_webhook_secret
PROXY_REVIEWER_USERNAME=revu-bot-reviewer
PROXY_REVIEWER_TOKEN=proxy_user_token

# Optional
ANTHROPIC_MODEL=claude-sonnet-4-5-20250929
ANTHROPIC_EXTENDED_CONTEXT=true
WEBHOOK_PROXY_URL=https://smee.io/your-url

# Optional: Valkey/Redis cache (recommended for discussion replies)
# When REDIS_URL is set, Revu uses Redis-backed compute cache; otherwise it falls back to in-memory.
# Example for docker-compose setup (service name "valkey"):
# REDIS_URL=redis://valkey:6379/0
# REDIS_DB=0
# REDIS_PASSWORD=
# REDIS_TLS=false

See .env.example for an example.

Choosing a provider (Anthropic or OpenAI)

Revu supports both Anthropic and OpenAI. You can select the provider either via config.json or an environment variable.

Option 1: config.json

{
  "promptStrategy": "line-comments",
  "thinkingEnabled": true,
  "llmProvider": "openai"
}

• llmProvider: "anthropic" (default) or "openai"

Option 2: Environment variable

You can also set the default provider via the LLM_PROVIDER env var:

# Allowed values: anthropic | openai
LLM_PROVIDER=openai

Precedence rules:

1. If config.json sets llmProvider, it takes precedence over the environment.
2. Otherwise, LLM_PROVIDER (when valid) overrides the built-in default.
3. If neither is set, Revu defaults to anthropic.

Environment variables per provider:

• Anthropic (default):

  • Required: ANTHROPIC_API_KEY
  • Optional: ANTHROPIC_MODEL (default: claude-sonnet-4-5-20250929)
  • Optional: ANTHROPIC_EXTENDED_CONTEXT=true to enable 1M context (beta API)

• OpenAI (official endpoint):

  • Required: OPENAI_API_KEY
  • Optional: OPENAI_MODEL (default: gpt-5)

Example OpenAI env:

OPENAI_API_KEY=your_openai_key
# Optional model override
OPENAI_MODEL=gpt-5

Running Revu

Local Development

# Dry-run review of a PR using the current local version of Revu
yarn review-pr https://github.com/owner/repo/pull/123
# Submit comments to GitHub after analysis
yarn review-pr https://github.com/owner/repo/pull/123 --submit

Production

# Local machine
yarn build
yarn start

Usage

1. Install Revu on your GitHub repositories
2. When a PR is opened, Revu automatically adds the proxy user as a reviewer
3. Click "Request review" from the proxy user to trigger code review
4. Revu analyzes the code and posts detailed feedback

For CLI usage and testing, see CLI Documentation.

Discussion replies

When a human user replies to a Revu inline review comment on a PR, Revu posts a concise threaded reply under the same review thread.

Behavior:

• Triggers only for replies (not new root comments) to Revu-marked comments.
• Requires the replying user to be an organization member (or at least a repo collaborator if org membership cannot be checked).
• Reuses the same review context (PR diff, modified files’ contents, coding guidelines, related issues) to ground the answer.
• Responses are concise; if a concrete fix is clear, a single GitHub suggestion fence is included.

Setup checklist:

• Enable the GitHub App event: “Pull request review comment”.
• Grant permission: “Organization members: Read”.
• Ensure PROXY_REVIEWER_USERNAME and PROXY_REVIEWER_TOKEN are configured.

Compute cache:

• Discussion replies are cached to avoid duplicate work on repeated deliveries.
• Cache key includes repo, PR, thread ids, reply content hash and commit SHA.
• Cache TTL defaults to 1 hour. Editing the user’s reply invalidates the cache automatically.

Configuration

Revu is configurable through a .revu.yml file in your repository root:

# Enable Extended Thinking
thinkingEnabled: true

# Custom coding guidelines
codingGuidelines:
  - 'Use descriptive variable names'
  - 'Add comments for complex logic'

# PR validation rules
validation:
  maxFilesChanged: 75
  maxDiffSize: 15000

# Branch filtering
branches:
  patterns:
    - '!**'
    - 'main'
    - 'release/*'

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

License

MIT License - see LICENSE file for details.