Nebula.AI - AWS Solution Architect Tool

Nebula.AI is an AI-powered co-architect that helps cloud engineers and architects design, validate, and deploy enterprise-grade AWS solutions at speed. It bridges human intent and cloud automation with intelligence, precision, and trust.

🚀 Features

• Three Intelligent Modes:
  • 🧠 Brainstorm Mode: Explore AWS services and best practices with AI-powered insights
  • 🔍 Analyze Mode: Get comprehensive analysis and intelligent recommendations for your requirements
  • ⚡ Generate Mode: Generate deploy-ready CloudFormation templates
• Real Strands Agents Integration: Uses actual Strands Agents SDK with MCP tool integration
• Mode-Based MCP Orchestration: Automatically selects and configures MCP servers based on selected mode
• Production-Ready Outputs: Real CloudFormation templates
• Intent-Based Analysis: Intelligent requirements analysis with keyword and intent detection

🏗️ Architecture

• /frontend → React + TypeScript + Tailwind CSS
• /backend → FastAPI + Strands Agents + AWS Core MCP Server
• /docs → product & prompt trackers
• /prompts → evolving prompt strategies
• /tests → automation tests

⚡ Quick Start

Prerequisites

• Python 3.11+ with AWS credentials configured
• Node.js 18+
• AWS Bedrock access (or Anthropic API key)

Running on Replit (Optional)

The application is pre-configured for Replit deployment:

1. Configure Environment Variables:

   • Edit backend/.env with your AWS credentials
   • Set AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION
   • Optionally set ANTHROPIC_API_KEY as fallback

2. Access the Application:

   • Frontend is available through Replit's webview (automatically opens)
   • Backend API: Check the backend workflow console for port info
   • API Docs: Add /docs to your Replit domain URL

3. Workflows:

   • Both frontend and backend start automatically
   • Frontend runs on port 5000 (webview)
   • Backend runs on port 8000 (console)

Local Setup

# Clone and setup
git clone <repository>
cd solutionbuilder_core

# Windows
run_dev.bat

# Linux/Mac  
./run_dev.sh

Access (Local)

• Frontend: http://localhost:5000
• Backend API: http://localhost:8000
• API Docs: http://localhost:8000/docs

Amazon Linux 3 Deployment

For production deployment on Amazon Linux 3 EC2 instances:

# Quick setup using automated script
cd backend
bash setup_amazon_linux3.sh

Key Requirements:

• Python 3.12+ (installed via script)
• Node.js 18+ (installed via script)
• Graphviz for diagram generation (installed via script)
• AWS credentials configured (IAM role or .env file)

Production Deployment:

• Use systemd services for automatic startup
• Configure firewall rules for ports 8000 (backend) and 5000 (frontend)
• Set up reverse proxy (nginx/Apache) for HTTPS in production

🔧 Configuration

1. AWS Credentials: Configure AWS CLI with Bedrock access, or set directly in backend/.env
2. Environment Variables:
   • Copy backend/env.example to backend/.env and update with your values
   • Required: AWS_ACCESS_KEY_ID, AWS_SECRET_ACCESS_KEY, AWS_REGION
   • Optional: ANTHROPIC_API_KEY (fallback if Bedrock unavailable), BEDROCK_MODEL_ID
3. MCP Servers: Automatically configured based on selected mode (see backend/config/mode_servers.json)
4. Performance: MCP servers are pre-installed locally for faster startup using uv tool install

Replit-Specific Notes (Optional)

• Frontend configured to run on port 5000 (Replit's webview port)
• API communication uses proxy (/api) to avoid CORS issues
• Both workflows start automatically when you open the Repl

📚 Documentation

• Product Tracker: See docs/product_tracker.md for change history
• Prompt Tracker: See prompts/prompt_tracker.md for prompt evolution
• Strands Agents: https://strandsagents.com/1.x/
• Core MCP Server: https://awslabs.github.io/mcp/servers/core-mcp-server

🎯 How It Works

1. Select Mode: Choose from Brainstorm, Analyze, or Generate mode
2. MCP Orchestration: System automatically selects and configures relevant MCP servers based on mode
3. Agent Execution: Strands Agents generate outputs using real AWS data and documentation
4. Export Results: Download CloudFormation templates

📡 API Endpoints

Core Endpoints:

• POST /brainstorm - AWS knowledge access for brainstorming
• POST /analyze-requirements - Enhanced requirements analysis
• POST /generate - Generate CloudFormation templates
• POST /follow-up - Handle follow-up questions with context
• GET /health - Health check endpoint

Streaming Endpoints:

• POST /stream-response - Stream brainstorm responses
• POST /stream-analyze - Stream analyze responses
• POST /stream-generate - Stream generate responses

Additional Endpoints:

• GET /mcp-pool-stats - MCP connection pool statistics