vision_QA_medgemma / README.md
codegood's picture
Initial commit: MedGemma medical VQA chat UI
cbee686
|
Raw
History Blame Contribute Delete
638 Bytes
---
title: MedGemma Medical VQA
emoji: 🩺
colorFrom: blue
colorTo: green
sdk: gradio
sdk_version: 6.20.0
app_file: app/chat_ui.py
pinned: false
---
# MedGemma Medical VQA
Multimodal medical Q&A chatbot. Upload a medical image (optional) and ask a
question; answers are grounded in NIH MedQuAD reference material via RAG,
generated by `google/medgemma-1.5-4b-it` served on AWS SageMaker (DJL-LMI +
vLLM, `ml.g5.xlarge`, scale-to-zero).
Requires these Space secrets: `AWS_ACCESS_KEY_ID`, `AWS_SECRET_ACCESS_KEY`,
`AWS_DEFAULT_REGION` (scoped to invoke the SageMaker endpoint and read/write
the S3 input/output/failures prefixes only).