Hearth

Inspiration

A drip, an unusual noise, or a flickering light leaves people with two questions: what should I do next, and what should I tell someone who can help? Hearth focuses on turning an uncertain household concern into a clearer conversation with a qualified professional.

What it does

Hearth is a working web simulation for the Alexa+ track. Describe a fictional home-maintenance concern, answer a few observation-based questions, and review a cautious next-step plan. The same observations become a contractor request you can edit, copy, or download. Hearth does not contact a contractor or book work.

Warning signs interrupt the ordinary flow. A simulated gas report skips routine questions and prioritizes leaving and calling for help from safety. Uncertain answers are supported, and incomplete safety information leads to professional-only guidance.

How it was built

HTML, CSS, and JavaScript provide the interface. One deterministic JavaScript rule engine powers the browser flow and the local integration endpoints. Node.js provides a dependency-free local HTTP server and test runner. The repository also contains a local MCP Streamable HTTP prototype using protocol revision 2025-11-25, exposing triage_home_issue and create_maintenance_plan. The static browser demo does not deploy that MCP server.

Codex assisted with the interface, shared planning logic, HTTP/MCP server, automated tests, standalone HTML build, documentation, and debugging. AI was used during development. The running app does not call a live AI model or Alexa+. This entry uses the hackathon's simulated Alexa+ experience path. It is independent, is not certified by Amazon, and does not claim live Alexa interoperability.

Challenges and accomplishments

The main challenge was keeping the plan and request consistent as answers change, while ensuring warning signs override routine maintenance guidance. The implementation invalidates stale plans, preserves deliberate request edits, handles uncertain answers, and keeps the simulation boundaries visible.

All 157 automated tests passed during submission preparation. Coverage includes the rule engine, HTTP/MCP transport, security boundaries, UI state with DOM doubles, and standalone artifacts. The actual browser walkthrough was then checked and recorded: the leak questions, resulting plan, editable subject, copied request, simulated transcript, and gas-warning route worked. Mobile behavior, real browser text-file downloads, and live Alexa interoperability remain unverified.

Try it

1. Download or clone the GitHub repository linked with this entry.
2. Open dist/hearth-demo.html in a modern browser. The standalone flow needs no account, server, API key, or installation.
3. Choose “A drip under the sink,” continue, and select “Under a sink or by a pipe,” “A slow, contained drip,” and “No visible electrical contact or ceiling sagging.”
4. Review the plan and edit the contractor request. Copy or download the unsent draft.
5. Start a new issue and choose the gas-safety demonstration to see the urgent warning route.
6. For the local server and integration checks, use Node.js 22 or later, run npm start, and follow the README's initialize/tools/list/tools/call examples. Run npm run check for syntax checks, tests, and the standalone build.

Use fictional data. This prototype cannot diagnose a fault or guarantee safety. Deterministic rules can miss hazards, misunderstand language, or produce false alarms. No real-world safety validation or user research has been performed. Do not use it in an actual emergency or expose the local development server publicly.

What's next

I’d like to turn this prototype into a working Alexa integration. That would require real-client interoperability checks, deployment authentication, remote hosting, and further safety and usability testing.

Built With

• codex
• css
• html
• javascript
• model-context-protocol
• node.js