CodePullAI

An intelligent AI assistant that automatically reviews pull requests to enhance code quality and security.

Comment

Inspiration

Our inspiration for CodePullAI came from a challenge every developer faces: the code review bottleneck.

As developers, we know that thorough code reviews are essential for maintaining high-quality, secure, and performant code. However, they are also time-consuming, can be subjective, and often slow down the development cycle.

Witnessing the rapid advancements in generative AI and participating in the AWS Serverless Hackathon, we saw a perfect opportunity. We were inspired to build a tool that could serve as an intelligent, unbiased assistant for developers, leveraging the power of serverless technology and cutting-edge AI like Amazon Bedrock to make code reviews faster, smarter, and more consistent.

What It Does

CodePullAI is an intelligent assistant that automates the code review process directly within GitHub.

When a developer creates or updates a pull request, CodePullAI springs into action:

  1. 🔔 Receives Real-Time Notifications Via a GitHub App, CodePullAI is instantly notified of changes in the repository.

  2. 🧠 Performs AI-Powered Analysis AWS Lambda processes the code diff and sends it to Amazon Bedrock via API, using the Claude model for deep contextual review.

  3. 🛡️ Provides Intelligent Feedback

It identifies:

  • Security vulnerabilities (XSS, injections)
  • Performance issues
  • Logical bugs
  • Maintainability problems
  • Best practice violations

  1. 💬 Posts In-line Comments It posts clear, contextual suggestions directly on the affected lines in the PR.

  2. 🧪 Manages Repositories A secure dashboard built with Next.js (authenticated via Supabase + GitHub OAuth) allows users to select which repositories the assistant should monitor.

How We Built It

We designed CodePullAI with a serverless-first to ensure scalability and cost-efficiency.

  1. 🔧 The Serverless Backend

Core services:

  • GitHub Webhook ➝ Triggers when a PR is opened or updated

  • Amazon API Gateway ➝ Receives the webhook securely

  • AWS Lambda:

    Validates the request signature

    Uses GitHub API to fetch PR data and code diff

    Sends the prompt to Amazon Bedrock (Claude)

    Receives feedback and posts comments via GitHub API

  1. 🖥️ Frontend & Authentication

    Built with Next.js, hosted on AWS Amplify

    Auth via Supabase GitHub OAuth for secure login and repo access

    Amazon DynamoDB used to store:

    User info

    Repository metadata

Challenges We Ran Into

• Complex Configuration: GitHub App credentials, IAM policies, Supabase OAuth, and Lambda environment variables all needed careful syncing.

• Webhook Debugging: Serverless debugging is non-trivial. We leaned heavily on CloudWatch Logs to trace execution paths.

• Prompt Engineering: Effective feedback required precise prompt crafting to Claude, instructing it to behave like a real reviewer with defined categories (e.g., security, performance).

• Security Management:

  • Sensitive secrets (GitHub private key) were stored in Lambda securely and never exposed to the frontend.
  • Webhook signature validation ensured secure API Gateway invocation.

Accomplishments That We’re Proud Of

  1. ✅ A Fully Functional End-to-End System: Git push ➝ AI code review ➝ GitHub PR comments

  2. ⚡ Pure Serverless Stack: No VMs, no containers — just Lambda, API Gateway, and DynamoDB

  3. 🔒 Smooth GitHub OAuth via Supabase: Easy onboarding for developers

  4. 🤖 Real, Actionable AI Integration: Generative AI that does useful work — not just a chatbot

What We Learned

  1. 🧩 Event-Driven Architecture: Real-world reactive systems are powerful and elegant

  2. ☁️ Deep Dive into AWS Serverless Stack: From IAM to API Gateway to Lambda triggers

  3. 🧠 Practical AI Integration: Effective use of Amazon Bedrock and Claude for real dev workflows

  4. 🔗 Multi-Platform Integration: We became pros at stitching GitHub + AWS + Supabase together securely

What's Next for CodePullAI

  • 🎛️ Customizable AI Reviewer Let users set preferences (e.g., strictness, security-focus, language standards)

  • 🧠 Deeper Code Context Analyze entire repo (not just diff) for more accurate insights

  • 🏢 Org & Team Support Manage PR settings and analytics across GitHub orgs

  • ⚙️ Infrastructure-as-Code (IaC) One-click deployment via AWS CDK or Terraform

  • 🌍 More Platform Support Extend to GitLab, Bitbucket, and others

Built With

  • amazon-bedrock
  • api-gateway
  • aws-lambda
  • dynamodb
  • github-api
  • next.js
  • supabase-auth
Share this project:

Updates