Inspiration
Repair advice from a single photo is often vague or overwhelming. I wanted a tool that guides users through a focused investigation.
What it does
FixTrace examines visible evidence, suggests likely causes, requests a useful follow-up photo when necessary, and produces a practical action plan.
How we built it
I created a two-stage AI pipeline separating visual inspection from diagnostic reasoning. The interface runs online and through Windows and macOS desktop wrappers, while the API key remains server-side.
Challenges we ran into
The main challenges were model timeouts, inconsistent responses, repeated follow-up requests, and adapting safety advice to different objects.
Accomplishments that we're proud of
We built a working investigation flow that compares multiple photos, tracks how theories change, adapts safety advice, and produces clear next steps across web and desktop platforms.
What we learned
We learned that separating visual observation from diagnostic reasoning creates more reliable results. Follow-up photos are most useful when they answer a specific unresolved question.
What's next for FixTrace
Next, we plan to support more repair categories, improve model speed and reliability, add user feedback, and professionally sign the desktop applications.
Built With
- ai
- cloudflare
- featherless
- glm
- kimi
- openai
- pyinstaller
- python
- pywebview
- qwen
- react
- sites
- typescript
- vinext
- vite
- webview2
- workers
Log in or sign up for Devpost to join the conversation.