module_vision.py

"""
File: module_vision.py
Description: A module for chat using image + text with a multimodal interface.
Author: Didier Guillevic
Date: 2025-05-08
"""

import gradio as gr
import vlm

def process(message, history):
    """Generate the model response given message and history
    """
    messages = vlm.build_messages(message, history)
    yield from vlm.stream_response(messages)

examples=[
    {
        'files': ['./passport_jp.png',],
        'text': 'Can you describe the image and present the information as a JSON document?'
    },
    {'text': 'How can we rationalize quantum entanglement?'},
    {'text': "Peux-tu expliquer le terme 'quantum spin'?"}
]

#
# User interface
#
with gr.Blocks() as demo:
    chat_interface = gr.ChatInterface(
        fn=process,
        description="Chat with text or text+image.",
        examples=examples,
        #cache_examples=False,
        stop_btn="Stop Generation",
        multimodal=True,
        type="messages"
    )