EchoForge 🛠️

Autonomous Hands-Free Edge Metrology AI Assistant for Aerospace Manufacturing
Submission for the Amazon Developer Hackathon: Build, Ship, Shape (2026)

---

🚀 Demo & Deployment Status

• Web App / Simulated Alexa+ Console: https://echoforge.aenox.me
• Live Edge MCP Server (Render): https://echoforge-mcp-fresh.onrender.com
  • Health check: GET https://echoforge-mcp-fresh.onrender.com/health
  • Streamable HTTP MCP Endpoint: POST https://echoforge-mcp-fresh.onrender.com/mcp (requires the server-side bearer token)
• Remediation branch: audit/echo-forge-remediation — security changes are implemented and pushed; deployment verification is pending.
• Primary Hackathon Track: Alexa+ (Self-Hosted Streamable HTTP MCP Server & Simulated Alexa+ Interface)
• Mini-Challenges Entered:
  1. AWS Builder: AWS Bedrock AgentCore (Amazon Nova Pro), AWS Amplify SSR, IAM Security Policies.
  2. Open Source: MIT Licensed repository with open-source MCP metrology bridge (see Open Source Submission Details).

---

🎯 The Problem & Our Solution

In high-precision aerospace manufacturing (turbine blades, engine flanges, casing bosses), coordinate-measuring machines (CMM) output complex legacy inspection files (e.g. .DMO / DMIS files) measured in fractions of a micron ($0.0001\text{ mm}$).

Cleanroom CNC machinists wearing protective gloves and face shields cannot operate keyboards or touchscreens while handling delicate components.

EchoForge bridges this gap:

1. Hands-Free Ambient Alexa+ Voice UX: High-res speech recognition with hands-free loop and real-time barge-in wake word detection ("Echo", "Alexa", "Stop", "Cancel").
2. Bedrock AgentCore Intelligence: Powered by Amazon Nova Pro to interpret tolerance zones and explain metrology readings.
3. MCP Metrology Bridge: Implements the MCP 2025-11-25 Streamable HTTP transport (/mcp) to read simulated .DMO inspection data and return measurements and tolerance context.
4. Dynamic Speech Synthesis: Browser voice synthesis reads the assistant's metrology summary aloud.

EchoForge is a hackathon prototype using simulated inspection data. It is not a machine-control system. Its repair tool returns a review notice and does not provide CNC settings. End-to-end model output is still being hardened; do not use generated output as a manufacturing work instruction.

---

🏗️ Architecture

flowchart LR
    A["Operator Voice / UI<br/>(Alexa+ / Web)"] -->|"Speech Input / Prompt"| B["Next.js 14 Dashboard<br/>(echoforge.aenox.me)"]
    B -->|"Autonomous Tool Orchestration"| C["AWS Bedrock AgentCore<br/>(Nova Pro Harness)"]
    C <-->|"Streamable HTTP (2025-11-25)<br/>or REST Fallback"| D["EchoForge MCP Server<br/>(Render Edge Node)"]
    D <-->|"Ingests & Parses"| E[("Legacy .DMO Factory<br/>Inspection Data")]
    C -->|"Metrology Summary & Review Notice"| B
    B -->|"Phonetic Voice Synthesis + Barge-in"| A

Loading

---

⚙️ Supported Aerospace Metrology Features

Inspection Feature	Nominal Dimension	Tolerance Zone	System Response
UPPER_FLANGE	$50.000\text{ mm}$	$\pm 2.0\text{ µm}$	Reports simulated reading and tolerance status
MAIN_BORE_HOLE	$15.000\text{ mm}$	$\pm 0.1\text{ µm}$	Reports simulated reading and tolerance status
TURBINE_BLADE_ROOT	$32.500\text{ mm}$	$\pm 1.5\text{ µm}$	Reports simulated reading and tolerance status
INLET_NOZZLE_RING	$78.200\text{ mm}$	$\pm 2.5\text{ µm}$	Reports simulated reading and tolerance status
EXHAUST_STATOR_BOSS	$24.000\text{ mm}$	$\pm 1.8\text{ µm}$	Reports simulated reading and tolerance status

---

📂 Repository Structure

.
├── frontend-dashboard/       # Next.js 14 App Router UI & AWS Bedrock AgentCore API
│   ├── src/app/
│   │   ├── api/agent/        # AgentCore InvokeHarnessCommand & MCP tool interceptor
│   │   ├── page.tsx          # Hands-free voice interface, barge-in wake word & dynamic TTS
│   │   └── globals.css       # TailwindCSS styling & responsive layout
│   └── package.json
├── mcp-server/               # Self-hosted MCP Server (MCP Spec 2025-11-25)
│   ├── server.py             # Streamable HTTP (/mcp), SSE (/sse), and REST endpoints
│   ├── factory_data/         # Live CMM .DMO simulated telemetry
│   ├── test_streamable_http.py # Automated test suite for MCP Spec 2025-11-25
│   └── requirements.txt      # Python dependencies (mcp, starlette, uvicorn)
├── planning/                 # Architecture, submission checklists, & friction logs
│   ├── HACKATHON_REQUIREMENTS.md
│   └── hackathon_overview.md
├── LICENSE                   # Open-source MIT License
└── README.md                 # Project documentation

---

🧪 Local Quickstart & Testing

📖 Looking for a detailed step-by-step developer walkthrough? See the full Local Setup & Troubleshooting Guide.

1. Edge MCP Server (Python 3.12+)

cd mcp-server
python -m venv .venv
source .venv/bin/activate
pip install -r requirements.txt

# Run the MCP transport and security regression checks
python test_streamable_http.py
python -m unittest test_security.py test_mcp_logic.py

# Launch local server (default: port 8000)
python server.py

2. Frontend Dashboard (Next.js 14)

cd frontend-dashboard
npm install

# Configure environment variables (.env.local)
# APP_AWS_REGION=us-east-1
# APP_AWS_ACCESS_KEY_ID=your_access_key
# APP_AWS_SECRET_ACCESS_KEY=your_secret_key

npm run dev
# Open http://localhost:3000 in your browser

---

⚡ Service Health Check

The health route is public; the MCP and tool routes require a bearer token configured only on the server:

curl -s https://echoforge-mcp-fresh.onrender.com/health

Expected Response:

{
  "status": "healthy",
  "service": "EchoForge Metrology Bridge",
  "mcp_protocol_version": "2025-11-25",
  "transports": ["streamable-http (/mcp)", "sse (/sse)"]
}

The MCP service accepts public metrology queries without an MCP_AUTH_TOKEN. The agent API still needs user authentication and production abuse controls; see the remediation friction log.

---

📄 License

This project is open-source and released under the terms of the MIT License.