VETASSIST AI: A Multimodal Chatbot for Veterinary Guidance

Abstract

Vet Assist AI is a multimodal chatbot designed to provide preliminary veterinary guidance by processing both textual queries and visual data, such as images of pet injuries or symptoms. The system follows a sequential pipeline where user input is first validated, and images are analyzed using the BLIP (Bootstrapping Language- Image Pre-training) model to generate descriptive captions. These captions are combined with the user's query and a safety-centric prompt to guide a Large Language Model (LLM) in generating responses. Each response includes a risk assessment (Low, Medium, High, Emergency), recommended immediate actions, monitoring guidance, and, when appropriate, the contact information for nearby veterinarians Built on Fast API, Vet Assist AI was evaluated on simulated pet health scenarios covering common issues. Results demonstrate that the system can consistently provide contextually appropriate, safe, and informative guidance, helping pet owners distinguish between minor issues and urgent situations while connecting them to local veterinary resources. This work highlights the feasibility of applying multimodal AI in sensitive domains like veterinary medicine and lays the foundation for future integration with clinical databases and decision support systems.

Ähnliche Arbeiten