Fine-tuned from google/medgemma-1.5-4b-it on Amod/mental_health_counseling_conversations for use as the generator model in a mental-health RAG assistant. Trained with a <context>...</context> prompt slot so retrieved passages can be injected directly.

Supportive, informational mental-health responses grounded in retrieved context. Not a diagnostic tool or crisis service.

This model can produce incorrect or incomplete clinical information. It must not be used as a substitute for a licensed mental health professional or emergency services. Deployers should keep a crisis-keyword safety layer (see the chat interface in the training notebook) in front of any user-facing deployment.

Use is governed by the Health AI Developer Foundations terms of use (inherited from the base model), plus the license of the training dataset.