AI Dev Logger

An Automated Code Documentation Platform.

An enterprise-grade, full-stack Automated Code Documentation platform. Automated Code Logger intercepts GitHub webhooks, semantically analyzes code diffs using a hybrid Cloud/Local AI architecture, and provides a Retrieval-Augmented Generation (RAG) chat interface for generating technical documentation.

Developed by: Prashant Mahto

---

Core Features

Hybrid AI Architecture (Cloud + Local)

• Cloud LLM (Gemini 2.5 Flash): Handles heavy natural language processing, generating human-readable commit summaries, and powering the RAG chat engine.
• Local Embeddings (Ollama + nomic-embed-text): Processes vector embeddings locally, saving cloud quotas and running at zero cost, perfectly mapping code diffs to a 768-dimensional space.

Semantic "Smart" Search

• Move beyond keyword matching. Powered by pgvector in PostgreSQL, the platform allows you to search your codebase using intent and meaning (e.g., "fixing database connection errors" will find commits touching application.properties even if the words don't match).

RAG-Powered AI Assistant

• Context-Aware Chat: A slide-out Gemini assistant that reads your project's recent commit history to answer specific questions about your codebase.
• 1-Click Documentation Generation: Ask the AI to write a weekly summary or explain a feature, and instantly export the response as a formatted Markdown (.md) or paginated PDF (.pdf) file.

Automated GitHub Integration

• Secure webhook listener that instantly captures push events, extracting the author, timestamp, commit message, and raw syntax-highlighted code diffs.

Enterprise-Grade Security

• Stateless Asymmetric Authentication: Secures backend API routes using modern Elliptic Curve (ES256) JWTs validated via Supabase's JWKS (JSON Web Key Set) public endpoints.
• Strict CORS Boundary: Backend enforces strict Origin policies, rejecting unauthorized domain requests.
• HMAC Signature Verification: Cryptographically verifies incoming GitHub webhooks using a user-defined secret to prevent payload spoofing.
• Input Sanitization: Utilizes Jakarta Validation to restrict payload sizes and prevent buffer/injection attacks.

---

Technical Stack

Frontend

• Framework: React (Vite)
• Styling: Tailwind CSS
• Icons: Lucide React
• Document Export: jspdf (PDF generation), standard Blob API (Markdown)

Backend

• Framework: Java (Spring Boot 3.5.x)
• AI Integration: Spring AI (spring-ai-openai shim for Gemini, spring-ai-ollama)
• Security: Spring Security (OAuth2 Resource Server)
• Database ORM: Hibernate Core (with hibernate-vector extension)

Database & DevOps

• Database: Supabase (PostgreSQL with pgvector extension)
• Containerization: Docker & Docker Compose (Multi-stage builds)
• Local AI Server: Ollama

---

Local Development Setup

1. Prerequisites

• Node.js & npm
• Java 21+ & Maven
• Docker Desktop (for running the local database/Ollama container)
• Supabase Account
• Google Gemini API Key

2. Environment Variables

Create a .env file in your backend directory:

GEMINI_API_KEY=your_gemini_api_key
SUPABASE_DB_PASSWORD=your_supabase_db_password

Create a .env file in your frontend directory:

VITE_SUPABASE_URL=your_supabase_project_url
VITE_SUPABASE_ANON_KEY=your_supabase_anon_key

3. Database Initialization

Run the following SQL commands in your Supabase SQL Editor

-- Enable vector extension
CREATE EXTENSION IF NOT EXISTS vector;

-- Create core tables
CREATE TABLE projects (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    user_id UUID NOT NULL,
    repo_name VARCHAR NOT NULL,
    github_installation_id VARCHAR,
    webhook_secret VARCHAR,
    created_at TIMESTAMPTZ DEFAULT NOW()
);

CREATE TABLE commit_logs (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    project_id UUID NOT NULL REFERENCES projects(id) ON DELETE CASCADE,
    commit_hash VARCHAR NOT NULL,
    author VARCHAR NOT NULL,
    message TEXT NOT NULL,
    raw_diff TEXT,
    ai_summary TEXT,
    embedding vector(768),
    commit_timestamp TIMESTAMPTZ NOT NULL
);

CREATE TABLE chat_messages (
    id UUID PRIMARY KEY DEFAULT gen_random_uuid(),
    project_id UUID NOT NULL REFERENCES projects(id) ON DELETE CASCADE,
    role VARCHAR NOT NULL,
    content TEXT NOT NULL,
    created_at TIMESTAMPTZ DEFAULT NOW()
);

4. Running the Application

Ensure Ollama is running locally and the model is downloaded (ollama pull nomic-embed-text).

Start the backend: mvn spring-boot:run

Start the frontend: npm run dev