app.py

import gradio as gr
import os
import requests
import base64

# 假设 libra_eval 在你的 python 包 libra.eval 中
from libra.eval import libra_eval

def generate_radiology_description(
    prompt: str,
    uploaded_current: str,
    uploaded_prior: str,
    temperature: float,
    top_p: float,
    num_beams: int,
    max_new_tokens: int
) -> str:
    """
    核心推理函数：
    1. 仅通过用户上传的图片获取图像文件路径
    2. 调用 libra_eval 来生成报告描述
    3. 返回生成的结果或错误消息
    """

    # 确保用户上传了两张图片
    if not uploaded_current or not uploaded_prior:
        return "Please upload both current and prior images."

    # 模型路径
    model_path = "X-iZhang/libra-v1.0-7b"
    conv_mode = "libra_v1"

    try:
        # 调用 libra_eval 进行推理
        output = libra_eval(
            model_path=model_path,
            model_base=None,  # 如果有必要，可指定基础模型
            image_file=[uploaded_current, uploaded_prior],  # 两张本地图片路径
            query=prompt,
            temperature=temperature,
            top_p=top_p,
            num_beams=num_beams,
            length_penalty=1.0,
            num_return_sequences=1,
            conv_mode=conv_mode,
            max_new_tokens=max_new_tokens
        )
        return output
    except Exception as e:
        return f"An error occurred: {str(e)}"

# 构建 Gradio 界面
with gr.Blocks() as demo:
    # 标题和简单说明
    gr.Markdown("# Libra Radiology Report Generator (Local Upload Only)")
    gr.Markdown("Upload **Current** and **Prior** images below to generate a radiology description using the Libra model.")

    # 用户输入：文本提示
    prompt_input = gr.Textbox(
        label="Prompt",
        value="Describe the key findings in these two images."
    )

    # 上传本地图像（Current & Prior）
    with gr.Row():
        uploaded_current = gr.Image(
            label="Upload Current Image",
            type="filepath"
        )
        uploaded_prior = gr.Image(
            label="Upload Prior Image",
            type="filepath"
        )

    # 参数调节
    with gr.Row():
        temperature_slider = gr.Slider(
            label="Temperature",
            minimum=0.1,
            maximum=1.0,
            step=0.1,
            value=0.7
        )
        top_p_slider = gr.Slider(
            label="Top P",
            minimum=0.1,
            maximum=1.0,
            step=0.1,
            value=0.8
        )
        num_beams_slider = gr.Slider(
            label="Number of Beams",
            minimum=1,
            maximum=20,
            step=1,
            value=2
        )
        max_tokens_slider = gr.Slider(
            label="Max New Tokens",
            minimum=10,
            maximum=4096,
            step=10,
            value=128
        )

    # 用于显示模型生成的结果
    output_text = gr.Textbox(
        label="Generated Description",
        lines=10
    )

    # 点击按钮时触发的推理逻辑
    generate_button = gr.Button("Generate Description")
    generate_button.click(
        fn=generate_radiology_description,
        inputs=[
            prompt_input,
            uploaded_current,
            uploaded_prior,
            temperature_slider,
            top_p_slider,
            num_beams_slider,
            max_tokens_slider
        ],
        outputs=output_text
    )

if __name__ == "__main__":
    # 启动 Gradio 应用
    # 将 share 设置为 True 以便在 Hugging Face Spaces 中分享
    demo.launch(share=True)