We just pushed a massive update to SonicFix, effectively completing our "Ear & Eye" pipeline. Here is what's new in the build:
- The "Gemini 3 First" Architecture We are now running on the bleeding edge. Our backend attempts to use gemini-3-flash-preview for every diagnosis to leverage its superior reasoning capabilities.
The Safety Net: Because preview models can be unstable (we hit a few 503s!), we built a custom Fallback Cascade. If Gemini 3 is busy, the system instantly degrades to gemini-1.5-flash without the user ever knowing. It’s the best of both worlds: Innovation + Reliability.
- New "Chat-Style" Interface We completely overhauled the UI. Instead of a static result screen, SonicFix now features a Session-Based Chat Interface.
Visual Context: You can see the photo you snapped of the engine/machine.
Audio Playback: You can replay the mechanical sound you recorded right inside the chat bubble.
Diagnostic Cards: The AI response is now rendered as a beautiful "Health Card" showing the Fault, Severity, and Estimated Cost (PKR).
- Under the Hood
Frontend: Flutter (Riverpod + Material 3)
Backend: Firebase Cloud Functions (Python)
AI: Multimodal Fusion (Image + Audio + Context)
We are now recording the final demo video. Fingers crossed for the submission
Log in or sign up for Devpost to join the conversation.