HomeSense Autopilot

🏡✨ HomeSense uses AI to explain smart-home failures, safely undo unwanted changes, and test fixes before they happen again! 🔄💜

Comment

Inspiration

Smart homes are getting smarter, but what happens when something goes wrong? And what happens when the software powering them fails?

We built HomeSense Autopilot around one idea: intelligence isn't just about preventing failure. It's about knowing how to recover—and when to ask for help.

What it does

HomeSense Autopilot combines a smart-home simulation with a human-supervised software delivery workflow.

  • Smart-home intelligence: Simulates connected devices, automations, incidents, and safe recovery.
  • AI-powered assistance: Uses Groq and internal MCP tools for natural-language interaction, with deterministic safeguards and fallback behavior.
  • GitLab CI/CD: Automatically runs quality checks, tests, dependency audits, secret detection, and SAST.
  • GitLab Duo: Investigates pipeline failures and escalates security-sensitive decisions for human review.

How we built it

We used React, TypeScript, Node.js, Express, SQLite, MCP, and Groq for the application.

We integrated the project with GitLab CI/CD, GitLab Duo, Semgrep, dependency auditing, and secret detection. The application runs independently on Render, with a fail-closed security gate protecting the delivery workflow.

Challenges we ran into

Our biggest challenge was getting the actual GitLab pipeline green.

We encountered authentication issues, historical synthetic credentials flagged by security scanning, and three vulnerable dependencies.

GitLab Duo investigated the failures and recommended human review. We then used Codex under human direction to implement narrowly scoped fixes without disabling our security protections.

Accomplishments that we're proud of

  • 110 automated tests passing—94 unit/integration and 16 browser tests.
  • Seven successful GitLab CI jobs.
  • Zero findings in the final dependency, secret-detection, and SAST reports.
  • Two genuine GitLab Duo investigations with human-supervised remediation.
  • An independently deployed, publicly accessible application.
  • All achieved using free resources.

What we learned

We learned that responsible AI isn't measured by how many decisions it makes independently.

Sometimes, the most intelligent action is recognizing uncertainty, preserving safeguards, and involving a human.

We also learned how to build auditable CI pipelines, handle real security findings, and turn failed builds into opportunities to improve software reliability.

What's next for HomeSense Autopilot

We want to expand recovery scenarios, improve agent-assisted diagnostics, introduce richer incident timelines, and explore approval-controlled remediation workflows.

Our long-term vision is simple:

A home that knows how to recover. A development pipeline that knows when not to act alone.

Built With

Share this project:

Updates

Private user

Private user posted an update —

HomeSense Autopilot — From Red Pipelines to Green!

Our GitLab CI/CD pipeline initially failed after detecting a historical synthetic credential and three vulnerable dependencies.

Instead of bypassing security checks, we used GitLab Duo to investigate the failures and recommend human review. With human supervision, Codex implemented targeted fixes, updated dependencies, and preserved our security safeguards.

Final result: 7 passing CI jobs, 110 successful tests, zero reported security findings, and a fully working live deployment!

The lesson? Responsible AI isn't just about fixing problems automatically—it's about knowing when to involve humans.

Log in or sign up for Devpost to join the conversation.

Submission history