Skip to main content

AI-powered code reviewer using OpenRouter LLMs

Project description

Iara - AI Code Reviewer 🧜‍♀️

Iara - AI Code Review Agent

🇧🇷 Leia em Português

Iara is an automated, project-agnostic, configurable code review tool designed to run in CI/CD pipelines or locally via CLI. It connects directly to the LLM provider of your choice — OpenRouter (free models), OpenAI, Google Gemini, or Anthropic Claude.


🧜‍♀️ Iara Code Review 🧪 Tests codecov PyPI - Version License: MIT


🚀 Features

  • Agnostic: Configure your project context (Tech Stack, Rules) via JSON.
  • Multi-Provider: Connect directly to OpenRouter, OpenAI, Google Gemini, or Anthropic Claude.
  • Smart Fallback: Automatically tries free models if the preferred one fails (OpenRouter only).
  • Rules-Based (Static): Identifies dangerous patterns instantly without spending tokens (e.g., GetComponent in loops in Unity).
  • LLM-Based (Intelligent): Uses AI to understand logic, security, and context, going beyond syntax.
  • GitHub + GitLab: Native integration with both platforms, with automatic comments on PRs/MRs.
  • Multi-Language Reviews: Configure the output language — reviews can be written in English, Portuguese, Spanish, French, and more.

🧠 Capabilities

Iara combines different types of analysis for a complete review:

Type What does it do? Does Iara cover it? How?
Static Analysis Finds bugs by reading code (fast). Yes Via Extensions (Regex) and LLM.
Linting Fixes style and formatting. Yes LLM can suggest Clean Code.
SAST Finds security flaws in code. Yes Primary focus on vulnerability detection.
Dynamic Analysis Finds bugs by running the app (slow). ❌ No Focus on fast CI/CD (Code Review).

What does it detect?

  1. Unity / Game Dev:

    • Use of slow APIs (Find, GetComponent) in critical loops (Update).
    • Excessive memory allocation (Garbage Collection).
    • Excess logging (Debug.Log) in final builds.
  2. Security (General):

    • Hardcoded credentials (Passwords, API Keys).
    • Injection vulnerabilities (SQL, Command).
    • Missing input validation.
  3. Code Quality:

    • Complex or confusing logic.
    • Exception handling errors.
    • Refactoring suggestions for readability.

📦 Installation and Setup

1. Install

pip install iara-reviewer

2. Configure (Interactive Setup)

iara init

The wizard guides you through 5 steps:

  1. Language — Choose the review output language (en, pt-br, es, fr, etc.)
  2. Provider — Choose your LLM provider: openrouter (default, free), openai, gemini, or anthropic
  3. API Key — Enter the key for the chosen provider (validated and saved to ~/.iara/config.json)
  4. Project — Name, tech stack, description
  5. Preferences — Focus areas (Security, Performance, etc.)

Done! Project config is saved at .iara.json.

3. Use

git diff main | iara

Check authentication

iara auth status

Manual setup (without wizard)

Set the provider and its key via environment variables:

# OpenRouter (default — free models available)
export OPENROUTER_API_KEY="sk-or-..."

# OpenAI
export IARA_PROVIDER="openai"
export OPENAI_API_KEY="sk-..."

# Google Gemini
export IARA_PROVIDER="gemini"
export GEMINI_API_KEY="AIza..."

# Anthropic Claude
export IARA_PROVIDER="anthropic"
export ANTHROPIC_API_KEY="sk-ant-..."

API key resolution priority: environment variable > global config (~/.iara/config.json).

From source (Development)

git clone https://github.com/felipefernandes/iara.git
cd iara
pip install -e .

⚙️ Project Configuration

iara init automatically creates .iara.json. You can also create it manually:

{
  "project": {
    "name": "My Project",
    "description": "Project description.",
    "tech_stack": ["Python"]
  },
  "review": {
    "focus_areas": ["Performance", "Security"],
    "ignore_patterns": []
  },
  "model": {
    "preferred": "google/gemini-2.0-flash-exp:free",
    "fallback_enabled": true,
    "provider": "openrouter"
  },
  "language": "en"
}

Supported providers and example models

Provider provider value Example models
OpenRouter (default) openrouter google/gemini-2.0-flash-exp:free, meta-llama/llama-3.2-3b-instruct:free
OpenAI openai gpt-4o, gpt-4.5-preview, o1
Google Gemini gemini gemini-2.5-flash, gemini-2.5-pro
Anthropic Claude anthropic claude-opus-4-5-20250929, claude-sonnet-4-5-20250929

Note: Smart fallback to free models is only available for OpenRouter. When using openai, gemini, or anthropic, set "fallback_enabled": false.

The language field controls the review output language. Supported values: en, pt-br, es, fr, de, ja, zh, ko, ru, or any language the LLM understands.

You can also override provider, model, and language via environment variables:

export IARA_PROVIDER="anthropic"
export IARA_MODEL="claude-sonnet-4-5-20250929"
export IARA_LANGUAGE="pt-br"

A ready-to-use example is available at iara-example.json.


🧠 Memory (RAG) [New]

Iara now supports a local Retrieval-Augmented Generation (RAG) system to provide context-aware reviews.

1. Install Dependencies

pip install iara-reviewer[rag]
# or
pip install lancedb sentence-transformers torch numpy

2. Index Your Codebase

Run this command in your project root to create the local vector index:

iara memory index

This will parse your code (extracting functions, classes, and calls) and store it in .iara/data/lancedb.

3. Review with Context

Just run the review command as usual. Iara will automatically use the memory to retrieve relevant context for the changed code.

git diff main | iara

4. Manage Memory

To clear the index:

iara memory clear

🏃 How to Use

Via Pipe (Git Diff)

git diff main | iara

Via Environment Variable

export PR_DIFF=$(git diff main)
iara

Scan Mode (Static Analysis)

iara --scan ./path/to/project

Forcing a Provider and Model

# Anthropic Claude
export IARA_PROVIDER="anthropic"
export ANTHROPIC_API_KEY="sk-ant-..."
export IARA_MODEL="claude-sonnet-4-5-20250929"
git diff | iara

# OpenAI GPT-4o
export IARA_PROVIDER="openai"
export OPENAI_API_KEY="sk-..."
export IARA_MODEL="gpt-4o"
git diff | iara

# Google Gemini
export IARA_PROVIDER="gemini"
export GEMINI_API_KEY="AIza..."
export IARA_MODEL="gemini-2.5-flash"
git diff | iara

🐙 GitHub Integration

Add Iara to your GitHub repository in 2 steps:

1. Configure the secret

Go to Settings > Secrets and variables > Actions > New repository secret and add the key for your chosen provider:

Provider Secret name
OpenRouter OPENROUTER_API_KEY
OpenAI OPENAI_API_KEY
Google Gemini GEMINI_API_KEY
Anthropic ANTHROPIC_API_KEY

2. Create the workflow

Create the file .github/workflows/iara-review.yml.

With OpenRouter (default, free models):

name: Iara Code Review

on:
  pull_request:
    types: [opened, synchronize]

permissions:
  pull-requests: write
  contents: read

jobs:
  review:
    runs-on: ubuntu-latest
    name: AI Code Review
    steps:
      - name: Checkout
        uses: actions/checkout@v4

      - name: Run Iara Code Review
        uses: felipefernandes/iara@main
        with:
          openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

With Anthropic Claude:

      - name: Run Iara Code Review
        uses: felipefernandes/iara@main
        with:
          provider: anthropic
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          model: "claude-sonnet-4-5-20250929"
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

With OpenAI:

      - name: Run Iara Code Review
        uses: felipefernandes/iara@main
        with:
          provider: openai
          openai_api_key: ${{ secrets.OPENAI_API_KEY }}
          model: "gpt-4o"
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

With Google Gemini:

      - name: Run Iara Code Review
        uses: felipefernandes/iara@main
        with:
          provider: gemini
          gemini_api_key: ${{ secrets.GEMINI_API_KEY }}
          model: "gemini-2.5-flash"
        env:
          GITHUB_TOKEN: ${{ secrets.GITHUB_TOKEN }}

Iara will automatically:

  • Review the Pull Request diff
  • Post a comment with the review result

All available inputs

- uses: felipefernandes/iara@main
  with:
    provider: "openrouter"                         # openrouter (default), openai, gemini, anthropic
    openrouter_api_key: ${{ secrets.OPENROUTER_API_KEY }}  # when provider=openrouter
    openai_api_key: ${{ secrets.OPENAI_API_KEY }}          # when provider=openai
    gemini_api_key: ${{ secrets.GEMINI_API_KEY }}          # when provider=gemini
    anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}    # when provider=anthropic
    model: "google/gemini-2.0-flash-exp:free"     # override model
    config_path: ".iara.json"                     # config path (default: .iara.json)
    post_comment: "true"                           # post comment on PR (default: true)
    language: "en"                                 # review language

🦊 GitLab Integration

1. Configure variables

Go to Settings > CI/CD > Variables and add:

  • The key for your provider (e.g., OPENROUTER_API_KEY, ANTHROPIC_API_KEY, etc.)
  • IARA_PROVIDER: the provider name (e.g., anthropic) — omit for OpenRouter default
  • GITLAB_TOKEN: Personal/Project Access Token with api scope (required for MR comments)

2. Add to .gitlab-ci.yml

stages:
  - review

iara_code_review:
  stage: review
  image: python:3.11-slim
  script:
    - apt-get update && apt-get install -y --no-install-recommends git curl
    - pip install iara-reviewer
    - git fetch origin $CI_MERGE_REQUEST_TARGET_BRANCH_NAME
    - export PR_DIFF=$(git diff origin/$CI_MERGE_REQUEST_TARGET_BRANCH_NAME...$CI_COMMIT_SHA)
    - REVIEW=$(iara 2>/tmp/iara_stderr.txt) || true
    - echo "$REVIEW"
    - |
      if [ -n "$REVIEW" ] && [ -n "$GITLAB_TOKEN" ]; then
        PAYLOAD=$(python3 -c "
      import sys, json
      review = '''$REVIEW'''
      body = '## 🧜‍♀️ Iara Code Review\n\n' + review + '\n\n---\n*Reviewed by Iara - AI Code Reviewer*'
      print(json.dumps({'body': body}))
      ")
        curl -s -X POST \
          -H "PRIVATE-TOKEN: $GITLAB_TOKEN" \
          -H "Content-Type: application/json" \
          -d "$PAYLOAD" \
          "${CI_API_V4_URL}/projects/${CI_PROJECT_ID}/merge_requests/${CI_MERGE_REQUEST_IID}/notes"
      fi
  allow_failure: true
  rules:
    - if: $CI_PIPELINE_SOURCE == "merge_request_event"

Iara will automatically:

  • Review the Merge Request diff
  • Post a comment with the review result on the MR

A complete template is available at gitlab-ci.yml.


🔧 Any CI (Jenkins, CircleCI, etc.)

pip install iara-reviewer

# OpenRouter (default)
export OPENROUTER_API_KEY="sk-or-..."
git diff main...HEAD | iara

# Anthropic Claude
export IARA_PROVIDER="anthropic"
export ANTHROPIC_API_KEY="sk-ant-..."
export IARA_MODEL="claude-sonnet-4-5-20250929"
git diff main...HEAD | iara

🧪 Tests

python -m unittest discover tests

📜 License

MIT

Project details


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

iara_reviewer-1.5.0.tar.gz (40.3 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

iara_reviewer-1.5.0-py3-none-any.whl (29.9 kB view details)

Uploaded Python 3

File details

Details for the file iara_reviewer-1.5.0.tar.gz.

File metadata

  • Download URL: iara_reviewer-1.5.0.tar.gz
  • Upload date:
  • Size: 40.3 kB
  • Tags: Source
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for iara_reviewer-1.5.0.tar.gz
Algorithm Hash digest
SHA256 9e3d15cb0a2fc78b5f3f28b798f1248c18b9fa1c87e519d4b044fe4f0c3b9c27
MD5 860f9323a54b5f06efc547481aefeb4c
BLAKE2b-256 256bb60416b7e7e83fbc2eda8ab4ec11c1c985de3c200c8cbbf29dae87d9c1a5

See more details on using hashes here.

Provenance

The following attestation bundles were made for iara_reviewer-1.5.0.tar.gz:

Publisher: publish-pypi.yml on felipefernandes/iara

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

File details

Details for the file iara_reviewer-1.5.0-py3-none-any.whl.

File metadata

  • Download URL: iara_reviewer-1.5.0-py3-none-any.whl
  • Upload date:
  • Size: 29.9 kB
  • Tags: Python 3
  • Uploaded using Trusted Publishing? Yes
  • Uploaded via: twine/6.1.0 CPython/3.13.7

File hashes

Hashes for iara_reviewer-1.5.0-py3-none-any.whl
Algorithm Hash digest
SHA256 d7390da244f30d495f98bd134adfdd0fb497863669f847bbed99059b4ef1aed4
MD5 45bd6719e3b35d5a4aedb716050dc1f7
BLAKE2b-256 8cc1fc46192ebb216cdb62fe5e3b79119666e5799a3a193f68efa76246e9aafd

See more details on using hashes here.

Provenance

The following attestation bundles were made for iara_reviewer-1.5.0-py3-none-any.whl:

Publisher: publish-pypi.yml on felipefernandes/iara

Attestations: Values shown here reflect the state when the release was signed and may no longer be current.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page