⚡ Azure AI Code Assistant

AI-powered Azure SDK coding assistant for VS Code with inline suggestions, RAG-backed retrieval, feedback collection, and companion product UI.

🔗 Quick Links

• Extension (VS Code Marketplace): https://marketplace.visualstudio.com/items?itemName=TEAMEKLAVYA.azure-ai-code-assistant&ssr=false#review-details
• Frontend Website: https://azure-ai-code-assistant.vercel.app/
• Video Demo: https://drive.google.com/drive/folders/1Zb8M3IUdBzpEJxNmIYs8VTnT2gJGGbvD?usp=sharing
• GitHub Repository: https://github.com/ArjunDivraniya/Doppelganger-March-2026-Eklavya

🌟 Overview

This repository contains a full-stack project:

• A VS Code extension that detects Azure coding context and generates inline suggestions.
• A backend API that analyzes context, retrieves relevant Azure docs, and generates completions via LLM.
• A webview app for in-editor suggestion display.
• A frontend landing website for demos and product presentation.

✨ Key Features

• ⚡ Copilot-style inline ghost text suggestions inside VS Code.
• 🧠 Azure-aware detection for JS/TS/C# (storage, cosmos, identity, keyvault, service-bus, etc.).
• 🔎 RAG pipeline using ChromaDB + embedding service for documentation grounding.
• 🧹 Suggestion cleaning and deduplication before display.
• 🛠️ Quick Fixes + automatic Azure import injection.
• ⭐ Feedback endpoint and storage for continuous improvement.
• 🧪 Mock/fallback behavior for local development without full cloud setup.

🏗️ Architecture

flowchart LR
		A[VS Code Editor] --> B[Extension Host]
		B --> C[Azure Detector + Context Builder]
		C --> D[API Service /suggest]
		D --> E[Backend Controller]
		E --> F[RAG Service + ChromaDB]
		E --> G[Prompt Builder]
		G --> H["LLM Service: Groq Llama"]
		H --> I[Response Processor]
		I --> B
		B --> J[Inline Suggestion UI]
		B --> K[Webview Panel]
		K --> L["Feedback API endpoint"]
		L --> M[(MongoDB)]

Loading

🧩 Repository Structure

Doppelganger-March-2026-Eklavya/
├── README.md
└── azure-ai-code-extension/
		├── backend/                  # Express API + RAG + feedback persistence
		├── docs/                     # Architecture and flow docs
		├── extension/                # VS Code extension host + webview + frontend
		├── scripts/                  # Cross-platform install/start scripts
		└── shared/                   # Shared constants and types

🔧 Core Modules

VS Code Extension (azure-ai-code-extension/extension/src)

• extension.ts: activation, provider registration, commands, webview messaging.
• inlineProvider.ts: inline completion generation pipeline.
• codeWatcher.ts: debounced typing watcher + manual trigger flow.
• azureDetector.ts: Azure context detection.
• contextBuilder.ts: payload builder for backend requests.
• apiService.ts: backend call, normalization, overlap removal, fallback logic.
• importFixer.ts and importInjector.ts: quick fixes and import automation.
• feedbackService.ts: feedback posting and lightweight intent detection.

Backend (azure-ai-code-extension/backend/src)

• server.js: Express app, middleware, route wiring, health check.
• controllers/suggestController.js: end-to-end suggestion pipeline orchestration.
• services/contextAnalyzer.js: sdk + intent detection.
• services/ragService.js: vector retrieval with optional sdk filtering.
• services/promptBuilder.js: prompt construction.
• services/llmService.js: Groq LLM call + mock fallback.
• services/responseProcessor.js: output cleanup.
• controllers/feedbackController.js: feedback validation + MongoDB persistence.

UI Apps

• extension/webview: React app rendered in VS Code panel.
• extension/frontend: React marketing/demo website for the project.

🚀 Quick Start

1. Install dependencies

From azure-ai-code-extension:

Windows PowerShell:

.\scripts\install-all.ps1

macOS/Linux:

./scripts/install-all.sh

2. Configure backend environment

Create azure-ai-code-extension/backend/.env from .env.example and set values:

PORT=3000
GROQ_API_KEY=your_key_here
GROQ_MODEL=llama-3.3-70b-versatile
OPENAI_API_KEY=your_openai_key_if_using_openai_embeddings
EMBEDDING_PROVIDER=local
EMBEDDING_DIM=384
CHROMA_DB_PATH=http://localhost:8000
MONGO_URI=mongodb://localhost:27017/azure-ai-assistant

3. Start backend + extension watch mode

From azure-ai-code-extension:

Windows PowerShell:

.\scripts\start-dev.ps1

macOS/Linux:

./scripts/start-dev.sh

4. Run extension in VS Code

1. Open azure-ai-code-extension/extension in VS Code.
2. Press F5 to launch Extension Development Host.
3. Ensure editor.inlineSuggest.enabled = true.
4. Open a JS/TS/C# file and start typing Azure SDK code.

🧪 Backend Commands

From azure-ai-code-extension/backend:

npm run dev           # start API with nodemon
npm run embed:chunks  # embed docs into ChromaDB
npm run test:rag      # run RAG smoke test
npm run test:filtering

🔌 API Endpoints

GET /health

Returns service health.

POST /suggest

Request:

{
	"language": "typescript",
	"imports": ["@azure/storage-blob"],
	"currentLine": "const blobServiceClient = new",
	"context": "previous code",
	"cursorPosition": { "line": 14, "character": 28 }
}

Response:

{
	"suggestion": "const blobServiceClient = new BlobServiceClient(...)"
}

POST /feedback

Request:

{
	"suggestion": "...",
	"rating": "positive",
	"sdkType": "blob-storage",
	"intent": "upload-data"
}

📚 Documentation Map

• Project architecture: azure-ai-code-extension/docs/architecture.md
• Extension flow: azure-ai-code-extension/docs/extension-flow.md
• Backend setup: azure-ai-code-extension/backend/BACKEND_SETUP_GUIDE.md
• RAG setup: azure-ai-code-extension/backend/RAG_PIPELINE_SETUP.md
• Extension analysis: azure-ai-code-extension/extension/EXTENSION_ANALYSIS.md
• Frontend website overview: azure-ai-code-extension/extension/frontend/WEBSITE_OVERVIEW.md

🤝 Team

Built by Team Eklavya for the March 2026 hackathon.

📄 License

See azure-ai-code-extension/extension/LICENSE.

☁️ Supported Azure Services

Service	JS/TS Import	C# Using
Azure Blob Storage	@azure/storage-blob	Azure.Storage.Blobs
Azure Cosmos DB	@azure/cosmos	Azure.Cosmos
Azure Key Vault	@azure/keyvault-secrets	Azure.Security.KeyVault.Secrets
Azure Identity	@azure/identity	Azure.Identity
Azure Service Bus	@azure/service-bus	Microsoft.Azure.ServiceBus
Azure Event Hubs	@azure/event-hubs	Microsoft.Azure.EventHubs
Azure AI Text Analytics	@azure/ai-text-analytics	(coming soon)
Azure Communication	@azure/communication-sms	(coming soon)
Azure Cognitive Search	@azure/search-documents	(coming soon)

---

🏆 Hackathon

• Built in: 24-48 hours
• Category: AI + Cloud + Developer Tools
• Key Innovation: Context-injected Azure knowledge (eliminates AI hallucinations)
• Unique Features list:
  • Intent detection from comments (// I need to upload to blob → full function)
  • Mock mode for zero-dependency development
  • Security-first: always suggests DefaultAzureCredential over connection strings
  • Multi-layer caching: session + Redis = 80%+ API call reduction
  • One boolean to switch from mock to real backend

---

👥 Team

Name	Role	Responsibility
Arjun	Extension Developer	codeWatcher, azureDetector, contextBuilder, apiService
Dhruvesh	UI Developer	Webview panel, React sidebar, accept/reject UI
Priy Mavani	Backend Developer	Express API, Azure OpenAI integration
Mayank Dudhatra	Backend Developer	RAG system, Redis caching, Azure, deployment, Researcher, Feedback system

---

---

📄 License

MIT License - feel free to use, modify, and distribute.

Built with ❤️ using Azure OpenAI, VS Code API, and TypeScript
Made for Hackathon 2025