VetThread-Voice by CAIOS

VetThread Voice is a realtime voice intake and veterinary handoff prototype for high-stress pet emergencies. Pet parents often describe events out of order, mix observations with questions, and omit timing or context. VetThread Voice uses AssemblyAI Realtime Speech-to-Text to capture the conversation, then applies an evidence layer that keeps structured observations tied to finalized owner-reported source turns. The product deliberately avoids diagnosis, treatment, dosing, prescription, prognosis, and unsupported causal claims. Instead, it focuses on capture, clarification, structure, provenance, safety routing, and professional handoff. For the hackathon demo, a simulated midnight feline emergency shows how the system moves from realtime speech to a structured handoff and then into a deterministic RED_ALERT_CRITICAL UI state when configured safety signals are detected. The commercial direction is B2B SaaS for veterinary clinics, emergency hospitals, and veterinary ecosystem partners through monthly or annual subscriptions. it captures spoken observations, preserves source evidence, asks bounded clarifying questions, and routes configured high-risk signals into a guarded emergency UI state without diagnosing or prescribing.