A tool-using agent: it plans, calls FDA labels, interaction scans, adverse-event and recall data and your cabinet, then checks its own answer for unverified numbers. Ask in English, Telugu or Hindi (set in Settings).
Pack mode reads brand, composition, batch and expiry. Pill mode describes shape, colour and imprint. Use a clear photo in good light. AI can misread — verify with a pharmacist.
Source: U.S. FDA drug labels (openFDA). Indian brands like Dolo or Crocin are mapped to their generic. The AI may only use label text; dose figures are verified against it.
Reminders fire while this page/app is open (browser limitation). Data stays on this device.
Enter 2+ medicines separated by commas. Each FDA label's interaction section is scanned for the others.
PharmAi Vision
Agentic assistant with tool calling (FDA labels, interactions, FAERS adverse events, recalls, cabinet, memory), self-check guard against unverified doses, vision scanning of packs and pills, medicine cabinet with reminders, voice in English/Telugu/Hindi, emergency escalation, and printable briefings.
The full production blueprint (CNN+ViT, LangGraph, Neo4j GraphRAG, Docker/K8s) is in the repository; this page is the free, GitHub-Pages edition.
Privacy: your Groq key, profile, cabinet and chat history are stored only in this browser. Prompts go directly to Groq; only public FDA lookups go elsewhere.
⚕ Educational information only — not medical advice, diagnosis or treatment. AI identification can be wrong. Never start, stop or change a medicine based on this tool; confirm with a pharmacist or doctor. Emergency: call your local number (US 911 · EU 112 · India 112) or US Poison Control 1-800-222-1222.