app.py

import os
import sys
import subprocess
import streamlit as st
import io 
import pypdfium2 
from PIL import Image, ExifTags
import logging

# 设置日志记录器
# logging.basicConfig(level=logging.INFO)
logging.basicConfig(level=logging.ERROR)
logger = logging.getLogger(__name__)


def resize_image_if_needed(pil_image, max_size_mb=1, max_edge_length=1024):
    """
    Detect the size of a PIL image, and if it exceeds 1MB or its long edge is larger than 1024 pixels,
    reduce its size to a smaller size.
    
    Args:
        pil_image (PIL.Image.Image): The input PIL image.
        max_size_mb (int): The maximum allowed size in megabytes.
        max_edge_length (int): The maximum allowed length of the long edge in pixels.
    
    Returns:
        PIL.Image.Image: The resized PIL image.
    """
    # Convert image to bytes and check its size
    img_byte_arr = io.BytesIO()
    pil_image.save(img_byte_arr, format='JPEG')
    img_size_mb = len(img_byte_arr.getvalue()) / (1024 * 1024)
    print(f"Image size: {img_size_mb} MB")
    
    # Check if the image size exceeds the maximum allowed size
    if img_size_mb > max_size_mb or max(pil_image.size) > max_edge_length:
        # Calculate the new size while maintaining the aspect ratio
        aspect_ratio = pil_image.width / pil_image.height
        if pil_image.width > pil_image.height:
            new_width = min(max_edge_length, pil_image.width)
            new_height = int(new_width / aspect_ratio)
        else:
            new_height = min(max_edge_length, pil_image.height)
            new_width = int(new_height * aspect_ratio)
        
        # Resize the image
        pil_image = pil_image.resize((new_width, new_height), Image.LANCZOS)
        
        # Convert the resized image to bytes and check its size again
        img_byte_arr = io.BytesIO()
        pil_image.save(img_byte_arr, format='JPEG')
        img_size_mb = len(img_byte_arr.getvalue()) / (1024 * 1024)
        
        # If the resized image still exceeds the maximum allowed size, reduce the quality
        if img_size_mb > max_size_mb:
            quality = 95
            while img_size_mb > max_size_mb and quality > 10:
                img_byte_arr = io.BytesIO()
                pil_image.save(img_byte_arr, format='JPEG', quality=quality)
                img_size_mb = len(img_byte_arr.getvalue()) / (1024 * 1024)
                quality -= 5
                
    return pil_image


def correct_image_orientation(pil_image):
    """
    自动检测PIL Image对象是否包含EXIF信息，如果包含则根据EXIF信息重新修改图片的朝向。
    
    :param pil_image: 输入的PIL Image对象
    :return: 返回修正后的PIL Image对象
    """
    try:
        # 获取EXIF信息
        exif = pil_image._getexif()
        if exif is not None:
            # 查找Orientation的EXIF标签编号
            for orientation in ExifTags.TAGS.keys():
                if ExifTags.TAGS[orientation] == 'Orientation':
                    break

            # 获取图片的朝向信息
            orientation_value = exif.get(orientation)
            print(f"Orientation value: {orientation_value}")

            # 根据朝向信息调整图片方向
            if orientation_value == 3:
                pil_image = pil_image.rotate(180, expand=True)
            elif orientation_value == 6:
                pil_image = pil_image.rotate(270, expand=True)
            elif orientation_value == 8:
                pil_image = pil_image.rotate(90, expand=True)

    except (AttributeError, KeyError, IndexError):
        # 如果没有EXIF信息或者没有朝向信息，跳过处理
        pass

    return pil_image


def clone_repo():
    # 从环境变量中获取 GitHub Token
    github_token = os.getenv('GH_TOKEN')

    if github_token is None:
        logger.error("GitHub token is not set. Please set the GH_TOKEN secret in your Space settings.")
        return False
    
    # 使用 GitHub Token 进行身份验证并克隆仓库
    clone_command = f'git clone https://{github_token}@github.com/mamba-ai/invoice_agent.git'
    repo_dir = 'invoice_agent'
    if os.path.exists(repo_dir):
        logger.warning("Repository already exists.")
        # 将仓库路径添加到 Python 模块搜索路径中
        # logger.warning(f"Adding {os.path.abspath(repo_dir)} to sys.path")
        # sys.path.append(os.path.abspath(repo_dir))
        return True
    else:
        logger.info("Cloning repository...")
        result = subprocess.run(clone_command, shell=True, capture_output=True, text=True)
    
    if result.returncode == 0:
        logger.warning("Repository cloned successfully.")
        repo_dir = 'invoice_agent'
        
        # 将仓库路径添加到 Python 模块搜索路径中
        sys.path.append(os.path.abspath(repo_dir))
        logger.warning(f"Adding {os.path.abspath(repo_dir)} to sys.path")
        return True
    else:
        logger.error(f"Failed to clone repository: {result.stderr}")
        return False



if clone_repo():
    # 克隆成功后导入模块    
    import invoice_agent.agent as ia
    # from invoice_agent.agent import load_models, get_ocr_predictions, get_json_result
    
    def open_pdf(pdf_file):
        stream = io.BytesIO(pdf_file.getvalue())
        return pypdfium2.PdfDocument(stream)


    @st.cache_data()
    def get_page_image(pdf_file, page_num, dpi=96):
        doc = open_pdf(pdf_file)
        renderer = doc.render(
            pypdfium2.PdfBitmap.to_pil,
            page_indices=[page_num - 1],
            scale=dpi / 72,
        )
        png = list(renderer)[0]
        png_image = png.convert("RGB")
        return png_image


    @st.cache_data()
    def page_count(pdf_file):
        doc = open_pdf(pdf_file)
        return len(doc)
    
    st.set_page_config(layout="wide")

    models = ia.load_models()

    st.title("""
    受領した請求書を自動で電子化 (Demo)
    """)

    col1, _, col2 = st.columns([.45, 0.1, .45])

    in_file = st.sidebar.file_uploader(
        "PDFファイルまたは画像:", 
        type=["pdf", "png", "jpg", "jpeg", "gif", "webp"],
    )

    if in_file is None:
        st.stop()

    filetype = in_file.type
    whole_image = False
    if "pdf" in filetype:
        page_count = page_count(in_file)
        page_number = st.sidebar.number_input(f"ページ番号 {page_count}:", min_value=1, value=1, max_value=page_count)

        pil_image = get_page_image(in_file, page_number)
    else:
        pil_image = Image.open(in_file).convert("RGB")
        pil_image = correct_image_orientation(pil_image)
    pil_image = resize_image_if_needed(pil_image)
    
    text_rec = st.sidebar.button("認識開始")

    if pil_image is None:
        st.stop()
    
    with col1:
        st.write("## アップロードされたファイル")
        st.image(pil_image, caption="アップロードされたファイル", use_column_width=True)
    
    # if 'json_predictions' in st.session_state:
    #     prev_json_predictions = st.session_state.json_predictions
    #     prev_excel_file_path = st.session_state.excel_file_path
    #     with col2:
    #         st.write("## 結果")
    #         # 提供下载链接
    #         with open(prev_excel_file_path, "rb") as file:
    #             st.download_button(
    #                 label="Download Excel",
    #                 data=file,
    #                 file_name="output.xlsx",
    #                 mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
    #             )
    #         st.write("解析後の内容:")
    #         st.json(prev_json_predictions)
        
    if text_rec:
        with col2:
            st.write("## 結果")
        
            # Placeholder for status indicator
            status_placeholder = st.empty()
        
            with st.spinner('現在ファイルを解析中です'):
                # Simulate model running time
                # time.sleep(5)  # Replace this with actual model running code
                # predictions = ia.get_ocr_predictions(pil_image, models)
                # json_predictions = ia.get_json_result(predictions)
                json_predictions = ia.get_json_result_v2(pil_image, models)
                logger.error(json_predictions)
                st.session_state.json_predictions = json_predictions
                
                # Convert JSON to Excel
                # excel_file_path = "output.xlsx"
                # st.session_state.excel_file_path = excel_file_path
                # ia.json_to_excel_with_links(json_predictions, excel_file_path)
            
                # After model finishes
                status_placeholder.success('ファイルの解析が完了しました!')

                # 提供下载链接
                # with open(excel_file_path, "rb") as file:
                #     st.download_button(
                #         label="Download Excel",
                #         data=file,
                #         file_name="output.xlsx",
                #         mime="application/vnd.openxmlformats-officedocument.spreadsheetml.sheet"
                #     )
            
                # Display the result
                st.write("解析後の内容:")
                st.json(json_predictions)
                # st.write(predictions)