# AI Code Review Tool - Implementation Plan

## Overview
A minimalist tool that receives Bitbucket webhooks for PR events, fetches code diffs, performs an AI review using LangChain, and emails a summary to the developer.

## Features
- **Multi-Repo Support**: One tool can handle webhooks from multiple repositories.
- **AI Review**: Focuses on Syntax, DRY (Don't Repeat Yourself), and Clean Code principles.
- **Asynchronous Processing**: Uses background tasks to prevent webhook timeouts.

## Proposed Changes

### 1. API Endpoint (app.py)
- `POST /webhook/bitbucket`: 
    - Extracts `workspace`, `repo_slug`, `pull_request_id`, and `author_email`.
    - Triggers a background review task.

### 2. Review Engine (scripts/ai_code_reviewer.py)
- **Diff Fetcher**: Calls Bitbucket API `GET /repositories/{workspace}/{repo_slug}/pullrequests/{id}/diff`.
- **LangChain Reviewer**: 
    - Prompt: "Review this diff for syntax, DRY violations, and quality."
    - LLM: OpenAI (GPT-4o) or Google (Gemini).

### 3. Notification (scripts/email_notifier.py)
- Formats the LLM output into a clean email.
- Sends via `mail_service` or SMTP.

## Setup Requirements
1. **Bitbucket App Password**: Create at Workspace level with `Pull Request: Read`.
2. **Environment Variables**:
    - `BITBUCKET_USERNAME`
    - `BITBUCKET_APP_PASSWORD`
    - `OPENAI_API_KEY` or `GOOGLE_API_KEY`
    - `SMTP_CONFIG` (if needed)