Pet AI: multilingual multi-agent triage assistant for cat owners
Multi-agent chatbot that helps cat owners judge how urgently their cat needs veterinary care, by text or voice in English and ten Indian languages, with a deterministic triage engine and a SOAP-format note for the vet.
Business Challenge
Cat owners often cannot tell whether symptoms need an emergency visit, a vet appointment soon, or home care, and language barriers make online advice harder to use. A language-model-only answer is risky for this decision because it can be inconsistent. The goal was a safe, auditable way to guide owners through an intake conversation and produce a clear urgency level.
Solution Delivered
A FastAPI backend with a small roster of Groq-hosted LLM agents (orchestrator, intake, conversation, safety, note and knowledge agents). Intake converts free-text symptoms into structured fields. Urgency (emergency, soon, home) is decided by a deterministic Python triage engine over a vet-reviewed knowledge base, not by an LLM. A Safety Agent can veto every drafted reply, and a Note Agent produces a SOAP consultation note. Voice input routes English to Groq Whisper and Indic languages to Bhashini ASR, with Bhashini speech output. A separate RAG knowledge base (sentence-transformers embeddings, 13 cat-care documents) answers general questions without influencing triage. The frontend uses Next.js and React.
Outcomes Achieved
Working end-to-end personal project: urgency decisions are deterministic and auditable, a Safety Agent reviews every reply, and the app supports English plus ten Indian languages (Hindi, Tamil, Telugu, Kannada, Malayalam, Marathi, Bengali, Gujarati, Punjabi, Odia) for voice and text. Not yet deployed to real users, so no usage metrics are claimed.
Measurable Business Outcome
Deterministic, vet-reviewed 3-level urgency triage (emergency / soon / home); 11 languages supported for voice and text; separate RAG knowledge base of 13 cat-care documents. Personal project, no production usage metrics yet.