ReviewRay

ReviewRay is a codebase-aware, automated AI code review engine designed to integrate with GitHub workflows. The application intercepts pull requests via webhooks, retrieves context from a vector database representing the codebase structure, generates reviews utilizing Large Language Models, and posts automated feedback directly onto the active code changes.

System Previews

Landing Page

Developer Dashboard

Account Settings and Billing

---

Core Features

• Contextual Codebase Intelligence: Resolves file structures and module imports to provide code reviews that understand project-wide architecture instead of analyzing files in isolation.
• Asynchronous Webhook Processing: Employs background task queues to process incoming GitHub pull request events without blocking main thread executions.
• Vector Semantic Search: Leverages vector embeddings to build an indexed map of repository code chunks, allowing fast retrieval of relevant code snippets.
• Automated Pull Request Integration: Automatically posts granular inline comments directly onto specific code lines in active pull requests.
• SaaS Subscription Lifecycle: Implements full billing lifecycles including upgrades, renewals, and cancellations.

---

Architecture and Data Flow

ReviewRay is built around a Retrieval-Augmented Generation (RAG) pipeline designed for software codebases. The data flow proceeds as follows:

[GitHub PR Event] ──> [Vercel Webhook Endpoint] ──> [Inngest Event Queue]
                                                             │
                                                             ▼
[Pinecone Index] <── [Fetch Code Context] <── [Parse PR Diff & Target Files]
       │
       ▼
[OpenRouter API] ──> [Assemble Prompts] ──> [Generate Automated Review]
                                                             │
                                                             ▼
                                                [Post Comments to GitHub]

1. Ingestion and Embedding: When a repository is synchronized, code files are retrieved, parsed into logical semantic chunks, processed into vector embeddings, and stored in a Pinecone index.
2. Event Interception: A dedicated GitHub App registers webhook subscriptions for active repositories. Opening or updating a pull request triggers an event payload.
3. Context Retrieval: The system reads the pull request diff, identifies files modified, and queries Pinecone to fetch semantically relevant code from the surrounding project.
4. Prompt Compilation: The system compiles a prompt comprising the pull request diff, the retrieved code context, and architectural guidelines.
5. Review Generation: The payload is sent to OpenRouter to run reasoning models.
6. Publishing Feedback: The structured review comment is formatted and posted back to the GitHub Pull Request using the Octokit client.

---

Technology Stack

• Core Framework: Next.js (App Router) utilizing React 19
• Language: TypeScript (Strict Mode)
• Database Layer: PostgreSQL managed via Prisma ORM
• Vector Database: Pinecone Client
• Queue Management: Inngest asynchronous job scheduling
• Authentication: Better Auth with GitHub OAuth integration
• Billing and Subscriptions: Razorpay Payment Gateway integration
• UI Styling: Tailwind CSS v4, Radix UI primitives, Lucide React, and Phosphor Icons

---

Environment Setup

To run the application locally, create a .env file in the root directory containing the following configuration:

# Database Configuration
DATABASE_URL="postgresql://user:password@host:port/dbname?sslmode=require"

# Better Auth Configuration
BETTER_AUTH_SECRET="your_32_character_session_signing_secret"
BETTER_AUTH_URL="http://localhost:3000"

# GitHub Auth and App Credentials
GITHUB_CLIENT_ID="your_github_oauth_client_id"
GITHUB_CLIENT_SECRET="your_github_oauth_client_secret"
GITHUB_APP_ID="your_github_app_id"
GITHUB_APP_NAME="your_github_app_name"
GITHUB_WEBHOOK_SECRET="your_github_webhook_secret"
GITHUB_APP_PRIVATE_KEY="-----BEGIN RSA PRIVATE KEY-----\nyour_github_app_private_key_escaped_newlines\n-----END RSA PRIVATE KEY-----"
NEXT_PUBLIC_GITHUB_PUBLIC_LINK="https://github.com/apps/your-app-slug"

# Vector Indexing
PINECONE_INDEX="your_pinecone_index_name"
PINECONE_API_KEY="your_pinecone_api_key"
PINECONE_HOST_NAME="your_pinecone_index_host_url"

# LLM Provider
OPENROUTER_API_KEY="your_openrouter_api_key"

# Payment Processor
NEXT_PUBLIC_RAZORPAY_TEST_API_KEY="your_razorpay_key_id"
NEXT_PUBLIC_RAZORPAY_TEST_KEY_SECRET="your_razorpay_key_secret"
NEXT_PUBLIC_RAZORPAY_PLAN_ID="your_razorpay_subscription_plan_id"
NEXT_PUBLIC_RAZORPAY_WEBHOOK_SECRET="your_razorpay_webhook_secret"

---

Installation and Execution

1. Clone the repository and install all dependencies:

   bun install

2. Generate the Prisma client interface:

   bunx prisma generate

3. Synchronize database schemas:

   bunx prisma db push

4. Start the local Inngest development server:

   bunx inngest-cli dev

5. Launch the Next.js development server:

   bun run dev

Open http://localhost:3000 to interact with the application.