2–5-doctor clinics don't have a scribe, a billing auditor, or a follow-up team. The doctor is all three. The day is the casualty.
Every other vet AI automates the exam room. Consilium owns the decision layer — before the consult, during it, and after it.
Everything below is built around one principle: the GLM is non-removable. Strip it out and you're left with a Telegram bot that says “call the clinic.”
Consilium reads unstructured visit notes, lab results, billing history, past prescriptions — and distills a trajectory the doctor can absorb in the 10 seconds before knocking on the exam-room door.
One dictation returns four artifacts: a SOAP note, a prescription with dosing, a billing checklist that flags items mentioned-but-unbilled in amber, and a staff to-do list. Doctor approves, edits, or rejects per card.
24–48h later, a Telegram message goes out. Replies come back in the owner's own words. The GLM classifies all-clear, monitor, or escalate — with differentials, confidence, recommended action, and a draft response. Doctor only sees what needs them.
Today's schedule, follow-up queue, and the month's KPIs — updated live via Supabase Realtime. Escalation cards appear without a page refresh. The doctor walks in and knows the state of the clinic in one glance.
One tap says the GLM was right or wrong. Rejections are logged and injected as few-shot examples into the next call — so the accuracy improves per clinic without any model retraining and without ever touching base-model weights.
After every visit, the patient's passport auto-updates at a public URL. Owners share it with boarding, travel, or any vet. Every scan surfaces Consilium. No extra doctor work, no paper vaccination cards.