DocDrift: AI-Assisted Documentation Drift Detector

DocDrift is an automated, AI-assisted system designed to detect and remediate documentation-code drift in software repositories. Whenever developers push code changes, DocDrift analyzes code diffs against existing prose documentation (READMEs, /docs directory, Wiki pages) using Google Gemini's semantic reasoning, generates specific patch suggestions, and presents them in a review dashboard for human approval before opening a fix pull request.

---

🚀 Features

• 🔍 Automated Drift Detection: Continuously monitors code changes via GitHub webhooks or REST triggers to spot outdated, inaccurate, or incomplete documentation.
• 💡 Actionable Patch Generation: Uses Google's Gemini LLM to generate precise, structured text replacements (oldText → suggestedText) along with AI reasoning.
• 🛡️ Human-in-the-Loop Oversight: No documentation is mutated automatically. Suggestions enter a review queue requiring explicit approval.
• 🎨 Modern Review Dashboard: Premium glassmorphic React dashboard for tech leads and reviewers to inspect side-by-side diffs and approve/reject suggestions with one click.
• 🗄️ Auditable Persistence: Tracks all drift reports, suggested fixes, and reviewer actions in PostgreSQL for full historical auditability.

---

🛠️ Technology Stack

Backend

• Framework: Spring Boot (Java 21)
• Data Access: Spring Data JPA, Hibernate
• Database: PostgreSQL (pgAdmin 4 integration) / H2 in-memory
• Networking: Spring WebFlux (WebClient for non-blocking GitHub & Gemini API calls)
• Security: Spring Security (CORS enabled for local frontend integration)

Frontend

• Framework: React (Vite)
• Icons: Lucide React
• Styling: Vanilla CSS (Custom dark-mode glassmorphic design system)

AI & Integrations

• LLM Engine: Google Gemini API (gemini-1.5-flash)
• Version Control: GitHub REST API & Webhook events

---

📐 Project Structure

Docdrift/
├── Docdrift/                 # Spring Boot Backend Project
│   ├── src/main/java/com/docdrift/Docdrift/
│   │   ├── config/           # Security & WebClient Configurations
│   │   ├── controller/       # REST Controllers (Webhooks, Repositories, Reviews)
│   │   ├── dto/              # Request/Response DTOs & Gemini payload models
│   │   ├── model/            # JPA Entities (DriftReport, DriftSuggestion, etc.)
│   │   ├── repository/       # Spring Data JPA Repositories
│   │   └── service/          # GitHub, Gemini LLM, & Diff Analysis logic
│   ├── src/main/resources/
│   │   └── application.properties
│   └── pom.xml
│
├── frontend/                 # Vite + React Dashboard Application
│   ├── src/
│   │   ├── components/       # Header, ReportList, SuggestionCard
│   │   ├── api.js            # Axios/Fetch integration with Spring Boot API
│   │   ├── App.jsx           # Main Dashboard Layout & Repo Tracking Modal
│   │   └── index.css         # Glassmorphism Design System
│   └── package.json
│
└── README.md

---

⚙️ Setup and Installation

Prerequisites

• Java 21 or later
• Node.js (v18+) & npm
• PostgreSQL running locally (or via pgAdmin 4)

---

1. Database Setup (PostgreSQL)

1. Open pgAdmin 4 or psql command line.
2. Create a new database named docdrift:

   CREATE DATABASE docdrift;

3. Update Docdrift/src/main/resources/application.properties with your PostgreSQL username and password:

   spring.datasource.url=jdbc:postgresql://localhost:5432/docdrift
   spring.datasource.username=postgres
   spring.datasource.password=YOUR_POSTGRES_PASSWORD

4. Set your API credentials in application.properties or environment variables:

   github.api.token=YOUR_GITHUB_PAT
   gemini.api.key=YOUR_GEMINI_API_KEY

---

2. Running the Backend (Spring Boot)

Navigate to the backend directory and launch the server:

cd Docdrift
.\mvnw.cmd spring-boot:run

The backend server will start on http://localhost:8080.

---

3. Running the Frontend Dashboard (React + Vite)

Open a new terminal window, navigate to the frontend folder, and start the Vite dev server:

cd frontend
npm install
npm run dev

The dashboard will be available at http://localhost:5173.

---

🧪 Testing the APIs via Terminal (curl)

1. Track a Repository

curl -X POST http://localhost:8080/api/repos \
  -H "Content-Type: application/json" \
  -d '{"url":"https://github.com/Rajan3103/Doc_Drift", "name":"Rajan3103/Doc_Drift"}'

2. Simulate a GitHub Webhook Push Event

curl -X POST http://localhost:8080/api/webhooks/github \
  -H "Content-Type: application/json" \
  -d '{
    "ref": "refs/heads/main",
    "repository": {"full_name": "Rajan3103/Doc_Drift"},
    "commits": [{"id": "sample-sha-12345", "message": "Updated function signatures"}]
  }'

3. Fetch Drift Reports

curl http://localhost:8080/api/reports

4. Approve a Suggestion

curl -X POST http://localhost:8080/api/suggestions/1/approve

---