pr-reviewer 0.4.0

AI-powered code review for GitHub & GitLab. Works with any LLM: GPT-4o, Claude, Llama, Gemini, Ollama. Free, open source.

AI Code Reviewer

AI-powered code review for GitHub pull requests.

One-line setup. Zero config. Works with any LLM.

Used by [X] developers · [Y] reviews completed · $0.002/review with Groq

Quick Start · /fix · Playground · Pre-Commit · Providers · /review · GitLab · VS Code · Self-Hosted

---

https://github.com/user-attachments/assets/demo-placeholder

Open a PR. Get AI review in 30 seconds. Inline comments with suggested fixes.

---

Why teams choose AI Code Reviewer

	AI Code Reviewer	CodeRabbit	GitHub Copilot	PR-Agent
Pricing	Free (bring your key)	$19/user/mo	$19/user/mo	Free (self-host)
LLM choice	Any (GPT, Claude, Llama, Gemini, Ollama)	Fixed	Fixed	GPT-4 only
Setup	30 seconds	5 minutes	Built-in	15 minutes
On-demand /review	Yes	Yes	No	Yes
Custom instructions	Yes .pr-reviewer.yml	Yes	No	Yes
Suggested code blocks	Yes (one-click apply)	Yes	No	Yes
Cost estimation	Yes (pre-review)	No	N/A	No
100% local option	Yes (Ollama)	No	No	No
Multi-language reviews	Yes (9 languages)	Yes	No	No
Auto-fix /fix	Yes (commits fixes to PR)	No	No	No
Web playground	Yes (try without install)	No	No	No
Pre-commit hook	Yes (block before PR)	No	No	No
VS Code extension	Yes	No	Built-in	No
GitLab CI support	Yes (native)	No	No	Yes
Self-hosted + RAG	Yes (agentic, AST-indexed)	No	No	Yes
Retry with backoff	Yes (all providers)	Unknown	N/A	No
Webhook idempotency	Yes (SHA dedup)	Unknown	N/A	No
Open source	MIT	No	No	Apache-2.0

---

Review summary

Inline comments with suggested fixes

---

Quick Start

1. Add the workflow (30 seconds)

Create .github/workflows/ai-review.yml:

name: AI Code Review

on:
  pull_request:
    types: [opened, synchronize, reopened]

permissions:
  contents: read
  pull-requests: write

jobs:
  review:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: mara-werils/ai-code-reviewer@v1
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

2. Add your API key

Go to Settings > Secrets > Actions and add your API key.

3. Open a PR

That's it. AI review lands in 30 seconds.

Want the cheapest option? Use Groq with Llama 3.3 — it's $0.002/review with a free API key.

---

Auto-Fix with /fix

Type /fix in any PR comment after a review — AI reads the review comments, generates fixes, and commits them directly to your branch.

You:  /fix
AI:   👀 Analyzing 3 issues from the review...
AI:   ## AI Code Fix
      Applied 3 fixes across 2 files.

      ### src/db.py
      - Fixed SQL injection by using parameterized query
      - Added input validation for user_id
        Commit: abc1234

      ### src/api.py
      - Added null check before accessing response.data
        Commit: def5678

How it works:

1. AI collects all [CRITICAL] and [WARNING] comments from the last review
2. For each file: reads the current content, generates fixed version via LLM
3. Commits each fix directly to the PR branch
4. Posts a summary of what was fixed and what was skipped

Requires contents: write permission. Add /fix support with the on-demand workflow.

---

Try It Online (No Install)

Paste any public GitHub PR URL and get an AI review instantly:

# Self-host the playground
git clone https://github.com/mara-werils/ai-code-reviewer.git
cd ai-code-reviewer
pip install . uvicorn starlette
export GROQ_API_KEY=gsk_...
export GITHUB_TOKEN=ghp_...
uvicorn playground.app:app --port 8000

# Or with Docker
docker build -t pr-reviewer-playground playground/
docker run -p 8000:8000 -e GROQ_API_KEY=gsk_... -e GITHUB_TOKEN=ghp_... pr-reviewer-playground

Open http://localhost:8000 — paste a PR URL, get a review. Share the link with your team.

---

Pre-Commit Hook

Review code before every commit. Catches issues before they reach the PR.

# .pre-commit-config.yaml
repos:
  - repo: https://github.com/mara-werils/ai-code-reviewer
    rev: v0.4.0
    hooks:
      - id: ai-code-review

# Install and run
pip install pre-commit
pre-commit install

# Now every git commit runs AI review on staged changes
git commit -m "feat: add new endpoint"
# AI Code Review: scanning 3 files (42 lines changed)...
# CRITICAL src/api.py:15
#   SQL injection vulnerability: user input passed directly to query
# Commit blocked: 1 issue(s) at or above threshold.

Configuration

Env Variable	Default	Description
PROVIDER	auto-detect	LLM provider
SEVERITY_THRESHOLD	critical	Block on: critical, warning, suggestion
MAX_COMMENTS	10	Max review comments
REVIEW_STYLE	concise	Review depth

Set SEVERITY_THRESHOLD=warning to also block on warnings. Use --no-verify to skip in a hurry.

---

On-Demand Review

Type /review in any PR comment to trigger a review on demand.

# Add to your workflow to enable /review command
on:
  pull_request:
    types: [opened, synchronize, reopened]
  issue_comment:
    types: [created]

jobs:
  auto-review:
    if: github.event_name == 'pull_request'
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: mara-werils/ai-code-reviewer@v1
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

  on-demand:
    if: >
      github.event_name == 'issue_comment' &&
      github.event.issue.pull_request &&
      startsWith(github.event.comment.body, '/review')
    runs-on: ubuntu-latest
    steps:
      - name: React to comment
        uses: actions/github-script@v7
        with:
          script: |
            await github.rest.reactions.createForIssueComment({
              owner: context.repo.owner,
              repo: context.repo.repo,
              comment_id: context.payload.comment.id,
              content: 'eyes'
            });
      - name: Get PR ref
        id: pr
        uses: actions/github-script@v7
        with:
          script: |
            const pr = await github.rest.pulls.get({
              owner: context.repo.owner,
              repo: context.repo.repo,
              pull_number: context.issue.number
            });
            core.setOutput('head_ref', pr.data.head.ref);
      - uses: actions/checkout@v4
        with:
          ref: ${{ steps.pr.outputs.head_ref }}
      - uses: mara-werils/ai-code-reviewer@v1
        env:
          OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

See the full example: examples/on-demand-review.yml

---

Supported Providers

Use any LLM. Switch providers with one line.

Provider	Model	Cost/review*	Setup
Groq	Llama 3.3 70B	~$0.002	GROQ_API_KEY (free tier)
Google	Gemini 2.0 Flash	~$0.003	GOOGLE_API_KEY
OpenAI	GPT-4o	~$0.05	OPENAI_API_KEY
Anthropic	Claude Sonnet	~$0.08	ANTHROPIC_API_KEY
Ollama	Any local model	$0.00	Setup guide
Azure OpenAI	GPT-4o	~$0.05	AZURE_OPENAI_API_KEY + API_BASE_URL
Any OpenAI-compatible	Any	Varies	OPENAI_API_KEY + API_BASE_URL

*Estimated for a ~200 line PR.

Provider examples

Groq (Llama 3.3 — nearly free, recommended to start)

- uses: mara-werils/ai-code-reviewer@v1
  with:
    provider: 'groq'
  env:
    GROQ_API_KEY: ${{ secrets.GROQ_API_KEY }}

Get a free API key at console.groq.com.

OpenAI (GPT-4o)

- uses: mara-werils/ai-code-reviewer@v1
  env:
    OPENAI_API_KEY: ${{ secrets.OPENAI_API_KEY }}

Anthropic (Claude)

- uses: mara-werils/ai-code-reviewer@v1
  with:
    provider: 'anthropic'
  env:
    ANTHROPIC_API_KEY: ${{ secrets.ANTHROPIC_API_KEY }}

Google (Gemini)

- uses: mara-werils/ai-code-reviewer@v1
  with:
    provider: 'google'
  env:
    GOOGLE_API_KEY: ${{ secrets.GOOGLE_API_KEY }}

Ollama (100% local, free)

- uses: mara-werils/ai-code-reviewer@v1
  with:
    provider: 'ollama'
    model: 'llama3.1:8b'
    api_base_url: 'http://your-server:11434/v1'

Any OpenAI-compatible API (LiteLLM, vLLM, etc.)

- uses: mara-werils/ai-code-reviewer@v1
  with:
    api_base_url: 'https://your-api.example.com/v1'
    model: 'your-model'
  env:
    OPENAI_API_KEY: ${{ secrets.YOUR_API_KEY }}

---

GitLab CI Integration

Works with GitLab merge requests via the CLI. Add to your .gitlab-ci.yml:

ai-code-review:
  stage: test
  image: python:3.11-slim
  rules:
    - if: $CI_PIPELINE_SOURCE == "merge_request_event"
  script:
    - pip install --quiet pr-reviewer
    - pr-reviewer review
        --platform gitlab
        --repo "$CI_PROJECT_PATH"
        --mr "$CI_MERGE_REQUEST_IID"
        --post
  allow_failure: true

Required CI/CD variables: GITLAB_TOKEN (api scope) + your LLM API key (OPENAI_API_KEY, GROQ_API_KEY, etc.)

Works with self-hosted GitLab too — set GITLAB_URL to your instance URL.

See the full template: examples/gitlab-ci.yml

---

Configuration

Action inputs

- uses: mara-werils/ai-code-reviewer@v1
  with:
    # LLM provider (openai, anthropic, groq, google, ollama)
    provider: 'openai'

    # Specific model (auto-selected per provider if empty)
    model: ''

    # Review style: concise, thorough, minimal
    review_style: 'concise'

    # Max comments per review (1-50)
    max_comments: '15'

    # Review comment language (en, zh, ja, ko, es, de, fr, ru, pt)
    language: 'en'

    # Custom instructions for your team
    custom_instructions: |
      - We use the Repository pattern
      - Flag any direct SQL queries in controllers
      - We handle PII — security is critical

    # Auto-label PRs (bugfix, feature, refactor, etc.)
    label_pr: 'false'

    # Suggest missing tests
    suggest_tests: 'true'

Per-repo config (.pr-reviewer.yml)

Drop this in your repo root for persistent config:

# .pr-reviewer.yml
review_style: thorough
max_comments: 20

custom_instructions: |
  - Our API uses FastAPI with Pydantic v2
  - All new endpoints must have OpenAPI docs
  - We're migrating from callbacks to async/await

ignore_paths:
  - "*.lock"
  - "generated/**"
  - "__snapshots__/**"

ignore_titles:
  - "WIP"
  - "DO NOT MERGE"

---

CLI

Review PRs locally or in any CI:

pip install pr-reviewer

# Review a GitHub PR
export GITHUB_TOKEN=ghp_...
export OPENAI_API_KEY=sk-...
pr-reviewer review --repo owner/name --pr 42

# Review with a specific provider
pr-reviewer review --repo owner/name --pr 42 --provider groq

# Review a local diff
git diff main..HEAD > changes.diff
pr-reviewer review --diff changes.diff

# Review and post back to GitHub
pr-reviewer review --repo owner/name --pr 42 --post

---

VS Code Extension

Review code directly in your editor — no PR needed.

Cmd+Shift+R  →  Review current file
Cmd+Shift+D  →  Review uncommitted changes
Right-click  →  Review selection

Install

Search "AI Code Reviewer" in the VS Code Extensions marketplace, or:

cd vscode-extension
npm install && npm run compile
# Then: Cmd+Shift+P → "Developer: Install Extension from Location..." → select vscode-extension/

Configure

Open Settings → search "AI Code Reviewer":

• Provider: openai, anthropic, groq, google, ollama
• API Key: your key (or use env vars)
• Review Style: concise, thorough, minimal

Issues appear in the Problems panel with severity levels (Error, Warning, Info, Hint).

Works with the same providers as the GitHub Action. $0.002/review with Groq, $0.00 with Ollama.

See the full docs: vscode-extension/README.md

---

Self-Hosted Mode

For teams needing full control, RAG-powered codebase understanding, and persistent analytics.

git clone https://github.com/mara-werils/ai-code-reviewer.git
cd ai-code-reviewer
cp .env.example .env  # Edit with your keys
docker-compose up -d

What you get

Feature	Description
Agentic review	LangGraph agent with 7 tools to investigate the codebase
AST-based indexing	Understands functions, classes, imports (Python AST + Tree-sitter for JS/TS/Go)
Hybrid retrieval	Semantic vector search + identifier matching + file neighborhood + RRF fusion
Cost tracking	Per-repository budgets and analytics dashboard
Feedback loop	Learns from developer reactions to comments
SSE streaming	Real-time review progress
Webhook idempotency	SHA-based deduplication prevents duplicate reviews
Retry + backoff	Handles rate limits, timeouts, and transient failures automatically

Architecture

graph LR
    A[GitHub Webhook] --> B[FastAPI]
    B --> C[Redis Queue]
    C --> D[Review Worker]
    D --> E[Classify PR]
    E --> F[Retrieve Context]
    F --> G[LangGraph Agent]
    G --> H{Tools}
    H --> I[search_codebase]
    H --> J[find_usages]
    H --> K[check_tests]
    H --> L[read_file]
    G --> M[Post Review]
    N[Indexer] --> O[(PostgreSQL + pgvector)]
    F --> O

---

FAQ

Is it free?

The tool itself is 100% free and open source (MIT). You pay only for LLM API calls. With Groq's free tier or Ollama, the total cost is $0.

Is my code sent to third parties?

Your code is sent to whichever LLM provider you choose. If you need full privacy, use Ollama with a local model — nothing leaves your network.

Does it work with private repos?

Yes. The GitHub Action uses your repository's built-in GITHUB_TOKEN, which has access to private repos.

Can I use it with GitLab / Bitbucket?

GitLab is fully supported! Use the CLI in your .gitlab-ci.yml with --platform gitlab. See the GitLab CI Integration section. Bitbucket support is on the roadmap.

How do I avoid noisy reviews?

1. Use review_style: minimal for less verbose reviews
2. Set severity_threshold: warning to skip info-level comments
3. Add custom_instructions to teach it your team's conventions
4. Use ignore_paths to skip generated files

Can I review in Chinese / Japanese / Korean / Spanish?

Yes! Set language: 'zh' (or ja, ko, es, de, fr, ru, pt).

How is this different from PR-Agent?

• Multi-LLM: Works with any LLM, not just GPT-4. Switch with one line.
• Simpler setup: One workflow file, no config needed.
• Cheaper: Groq at $0.002/review vs GPT-4 at ~$0.10/review.
• Cost estimation: Know the cost before running a review.
• Better reliability: Retry with exponential backoff, webhook idempotency.

---

Badge

Show that your project uses AI code reviews:

[![AI Code Review](https://img.shields.io/badge/AI-Code%20Review-blueviolet)](https://github.com/mara-werils/ai-code-reviewer)

---

Roadmap

• Multi-LLM support (GPT, Claude, Llama, Gemini, Ollama)
• Inline comments with suggested fixes
• On-demand /review command
• Self-hosted mode with RAG
• CLI tool
• Cost estimation
• GitLab integration
• Bitbucket integration
• PR chat — ask questions about the PR
• Auto-fix /fix — AI commits fixes directly to your PR branch
• Learning from feedback
• VS Code extension
• JetBrains plugin
• Slack/Discord notifications

---

Contributing

Contributions are welcome! See CONTRIBUTING.md for guidelines.

git clone https://github.com/mara-werils/ai-code-reviewer.git
cd ai-code-reviewer
pip install -e ".[dev]"
pytest tests/ -v
ruff check .

---

License

MIT — use it however you want.

---

If this saves you time, give it a star. It helps others find the project.

Report Bug · Request Feature · Discord · Discussions