FixQuest AI

Turn everyday home problems into interactive repair adventures.

Comment

Inspiration

Home maintenance can feel like a language nobody taught us. A dripping tap or a sticking door can trigger guesswork before we understand the problem. FixQuest AI makes those first observations approachable through a living-room learning game where people can pause, discuss a choice, and build confidence together.

What it does

Explore three illustrated adventures: The Dripping Tap, The Slow Sink, and The Sticking Door. Each mission includes five challenges, contextual hints and explanations, incorrect-answer feedback, an interactive puzzle, and a completion badge. Learners sequence safe observations or align a virtual door, then earn XP and improve their best scores.

The interface uses D-pad, Select, and Back. Original room illustrations, large typography, bright focus indicators, optional system narration, replay, and locally saved progress support a television-first experience. No account, backend, subscription, or API key is required.

QuestGuide is a deterministic Smart Learning Guide using curated, context-aware content. It works offline and is not live generative AI. This is an educational simulation, not professional repair advice or a diagnostic service. It teaches observation, safe limits, and when to seek qualified help.

How we built it

We built a native Android application targeting Fire OS with Java, Android Views, Canvas, local JSON, SharedPreferences, and Android TextToSpeech. Native focusable controls support directional navigation. Original procedural graphics render the house, rooms, toolkit, and badge without downloaded art.

Correct decisions earn 20 XP, mission completion adds 50 XP, and a perfect run adds 30 XP. Only each mission's highest completed score contributes to the total, preventing repeated play from inflating progress. Maximum total: 540 XP. The app requests no permissions and calls no backend or model API. Codex assisted with implementation, content, testing, and submission preparation.

Challenges we ran into

The main challenge was making a practical subject work with directional controls. Short choices and small puzzles keep interaction simple while encouraging observation and reasoning. Clear educational boundaries were equally important: similar symptoms can have different causes, so the game describes selected models and reinforces when to stop.

The available Amazon device was a Fire tablet, not a Fire TV. We used it for compatibility and D-pad input testing while keeping target-device validation outstanding.

Accomplishments that we're proud of

Three complete missions, original illustrations, interactive puzzles, contextual guidance, optional narration controls, persistent XP, badges, and replay are implemented. An automated device test passed 95 Select actions using actual D-pad key events, covering retries, hints, sequences, alignment, scoring, and Back navigation. The APK built successfully, lint reported zero errors, and an extracted portable source archive built successfully in a fresh directory.

What we learned

Carefully scoped, curated content can support a useful offline learning coach. Immediate explanations and a chance to retry create a productive learning loop. Television navigation makes hierarchy concrete: the current question, focused choice, and next action must be clear. Compatibility testing must also be distinguished from validation on the intended target device.

What's next

Validate on an actual Fire TV or accepted simulator, test a physical remote, and verify offline narration on the target device. Then expand the mission library with content reviewed by qualified practitioners, add localisation and adjustable text size, introduce more puzzles, and test with families and first-time renters. An optional generative coach could later use a secure backend with strict scope, clear labelling, and offline fallback.

Current validation status

Built and tested on an Amazon KFTRWI Fire tablet running Android API 28, including D-pad input and a 16:9 layout review. Actual Fire TV hardware or accepted Fire TV/Vega simulator validation has not been completed. The available 2:20 rehearsal is explicitly labelled tablet footage. Android narration controls are implemented, but audible TextToSpeech output was not independently verified.

Source and sideload build

Repository: https://github.com/lanray07/FixQuest-AI The submission folder contains the demo APK, source archive, test evidence, screenshots, and labelled tablet rehearsal. The APK is debug-signed for sideload evaluation, with minimum API 22 and no runtime permissions.

Built With

Share this project:

Updates

Submission history