This is a submission for the Gemma 4 Challenge: Build with Gemma 4
What I Built
This system allows Field Technicians to enter relevant problem statements to the system which the AI System performs diagnosis on using Gemma4 to output resolution plan and likely cause.
Demo
Code
FieldFix Gemma
FieldFix Gemma is an offline-first repair copilot for field technicians. It combines local retrieval over equipment manuals with Gemma 4 reasoning to turn rough field notes, readings, and optional photos into a safe diagnostic checklist and dispatch-ready work order.
Why Gemma 4 4B
This project intentionally targets the small Gemma 4 model class, especially the 4B model.
- Field technicians often work with weak connectivity.
- Repair notes, measurements, and manuals are private operational data.
- The task is narrow enough for a small local model when paired with retrieval.
- Low latency matters more than open-ended general intelligence at the point of repair.
The app keeps retrieval cheap and local, then asks Gemma 4 to do the real judgment: rank causes, apply safety constraints, decide the first checks, and produce structured work-order JSON.
Architecture
-
server.js: static app server and/api/diagnoseendpoint. -
src/diagnosisEngine.js: local manual retrieval, prompt construction, validation, and…
How I Used Gemma 4
I used Gemma4 4b latest model. The Frontend prompts the model with relevant data which ingests the data and the symptoms and performs relevant analysis and throws back to the UI to indicate the resolution and fix. This can speed up resolution of complex scenarios and assist technicians to cover more repairs per day or be determinate in their solution.
Top comments (0)