How DiaCare AI was built
A wrapper-model architecture for a health assistant: Arka Helix 1, a Cloudflare Worker, a 6,500-character system prompt, and local-first health profiles.
DiaCare AI is built by a single-digit team. Here is how it works.
The model
Arka Helix 1 — a wrapper model around a high-quality open-weights foundation, tuned for Indian context, diabetes education, and conservative safety. The system prompt is ~6,500 characters and includes explicit refusal patterns for diagnosis and prescription.
The backend
A single Cloudflare Worker. Receives the chat messages array, adds the model selection, forwards to the inference layer, streams the response back. Stateless, low-latency from anywhere especially India, generous free tier.
Health profile injection
User health profile lives in browser localStorage — not on our servers. On every chat call, the profile is appended to the system prompt under a clear USER HEALTH PROFILE tag. This lets Helix 1 personalise every response without us ever holding sensitive health data server-side.