SI-MAPPER is an AI Agent built using the Google ADK and Ashrae 223P standard. It transforms multimodal HVAC data into a semantic virtual twin, integrating BACnet, Modbus, and external databases for advanced diagnostics and optimization.

🚀 Quick Start

The project is structured with a Next.js frontend that orchestrates the backend services.

1. Prerequisites

• Node.js & pnpm
• uv (Python package manager)

2. Setup & Run (Development)

Navigate to the mapper folder to install dependencies and start the full system in development mode:

cd mapper
pnpm install
pnpm dev

This single pnpm install will automatically synchronize the Python environments for the Agent and MCP Server as well.

3. Setup & Run (Production/Optimized)

For a faster, more optimized experience, you can build the frontend and run the production server:

cd mapper
pnpm build
pnpm start

This will:

1. Build: Optimize and minify the Next.js frontend assets.
2. Start: Launch the production frontend along with the Agent and MCP Server backend processes.

---

🛠️ Independent Control

If you need to work on specific components independently:

Component	Path	Tool	Command
Frontend	/mapper	pnpm	pnpm dev:ui
Agent	/agent	uv	uv run main.py
MCP Server	/mcp_server	uv	uv run server/main.py

Tech Stack

• Agent Framework: Agent Development Kit
• Frontend: React
• Backend for the Agent: FastAPI
• Connection between the Agent and the Frontend: CopilotKit
• MCP Server: FastMCP
• Containerization: Docker
• Container Orchestration: Docker Compose
• Ontology Standard: Ashrae 223P
• Graph Database: Neo4j