Increase max_tokens to 1024 and improve prompts for detailed medical responses

Switch to llama-3.1-8b-instant (mixtral was deprecated)

Fix Groq API model name and improve error handling

Replace towhee milvus-client with direct MilvusClient search for Zilliz Serverless

Use MilvusClient API for Zilliz Serverless compatibility

Handle Milvus has_collection exception for Zilliz Cloud

Fix Milvus collection initialization and Towhee connection

Deploy VitalSync AI

initial commit