# AI Code Review Service

A FastAPI-based service that automatically reviews Bitbucket Pull Requests using AI (Google Gemini 1.5 Flash) and provides feedback via API or email.

## How it Works
1. **Webhook**: Bitbucket sends a POST request to this service whenever a Pull Request is created or updated.
2. **Diff Fetching**: The service fetches the code diff for the Pull Request from Bitbucket's API.
3. **Filtering**: The diff is filtered to only include relevant code files (e.g., `.py`, `.js`, `.php`). Non-code files like `.md` or `.env` are excluded to save costs.
4. **AI Analysis**: The filtered diff is sent to Google Gemini with a custom prompt focused on Syntax, DRY principles, Clean Code, Performance, and Security.
5. **Output**: The generated review is returned directly via API (for manual requests) or sent to the PR author via email (for webhooks).

## Setup Instructions

### 1. Environment Configuration
Create a `.env` file in the root of `code_review_service/` (refer to `.env.example`):
- `BITBUCKET_EMAIL` & `BITBUCKET_TOKEN`: Your Bitbucket email and App Password.
- `GEMINI_API_KEY`: Your Google Gemini API key.
- `GEMINI_MODEL`: Model name (default: `gemini-2.5-flash`).
- `FALLBACK_RECIPIENT`: Email address to use if the PR author's email is not found.

### 2. Configure Bitbucket Webhook
Follow these steps to connect your repository to this service:
1. Go to your repository on Bitbucket.
2. Click on **Repository settings** (bottom left).
3. Select **Webhooks** under the Workflow section.
4. Click **Add webhook**.
5. Fill in the details:
   - **Title**: AI Code Review
   - **URL**: `https://api-service-uat.creditfair.in/webhook/bitbucket` (or your appropriate base URL)
   - **Status**: Active
   - **Triggers**: Choose **Repository events** -> **Pull Request** -> **Created** and **Updated**.
6. Click **Save**.

### 3. Manual Review via API (Staging/Testing)
You can trigger a review manually without setting up a webhook by calling the review endpoint.

**Endpoint**: `POST /api/v0/review`  
**Body**:
```json
{
  "url": "https://api.bitbucket.org/2.0/repositories/{workspace}/{repo_slug}/pullrequests/{pr_id}/diff"
}
```

**Testing via Swagger UI**:
1. Run the app: `python app.py`
2. Open `http://127.0.0.1:8006/docs` in your browser.
3. Use the **Try it out** button on the `/api/v0/review` endpoint.

### 3. Local Development / Testing
```powershell
# Activate venv
.\venv\Scripts\activate
# Install dependencies
pip install -r requirements.txt
# Run service
python app.py
```

### 4. Deployment
- This service needs to be accessible from the internet for Bitbucket to send webhooks.
- Use a tool like `ngrok` for local testing or deploy to a public server (AWS, etc.).
- Ensure port `8006` (or your configured port) is open in the firewall.

## Project Structure
- `app.py`: FastAPI application and webhook endpoint.
- `src/utils/bitbucket_client.py`: Bitbucket API interactions.
- `src/utils/ai_reviewer.py`: LangChain and OpenAI logic.
- `src/utils/email_notifier.py`: Email sending logic using `core_utilities`.

## Cost Optimization
- **Model**: Uses `gemini-2.5-flash` which is highly cost-effective with a large context window.
- **Diff Only**: Only the code changes (diff) are sent to the AI, not the entire files.
- **File Filtering**: Only specific extensions ([.py, .js, .php, .html, .css, .ts]) are processed.
- **Concise Response**: The prompt enforces a short, high-impact summary to reduce token usage and improve readability.