Snap a photo or just ask. Ingestics analyses every ingredient, runs deep web research, rates safety, and shows the result in beautiful charts — like ChatGPT, but for your health.
Active ingredients, mechanism of action, dosage, side effects, interactions and patient-friendly summary.
Decodes the ingredient list, flags additives, computes a 0-10 health rating and tags healthy / unhealthy.
Macro splits, side-effect frequency bars and radar profiles you can read at a glance — when they help.
Just upload the package or pill blister. The agent reads the label, with OCR fallback if needed.
Cross-references each query with live web data + your own past chats so answers stay current and personal.
Auth + chats live in Supabase with row-level security. Images live on Cloudflare R2. Only you see them.
Upload a photo of the medicine blister or food packaging — or simply type your question.
The agent reads the image, classifies it as medicine, food, or a general health question.
It runs live web search + recall on your past chats and reasons over the result.
You get a flexible answer — sometimes a card with charts, sometimes a friendly paragraph.
Authentication and chat history live in Supabase under row-level security. Image uploads land in Cloudflare R2. Every inference call is gated, observed, cached and rate-shaped through Cloudflare AI Gateway — so your queries stay fast and private.
Each user only ever reads their own conversations and uploads. Server keys never reach the browser.
Runs in 300+ Cloudflare data centres so you get sub-100ms latency on every continent.
Live multi-source web search feeds the agent context, so answers reflect current product information.
No. Ingestics is an AI assistant — always consult a licensed healthcare professional before changing any medication.
Chats live in Supabase with row-level security. Image uploads live in your scoped Cloudflare R2 bucket. Only you can read them.
Yes — the agent handles 30+ languages on packaging, and replies in the exact language you write to it in.
The agent runs an OCR fallback on the server so even partial text from the wrapper is enough to identify the product.