# 核心依賴庫(kù)
pip install pdf2image pillow requests glob3 pathlib

import glob
import os
from typing import List

def get_all_pdf_in_dir(pdf_dir: str) -> List[str]:
    """
    獲取指定目錄下所有PDF文件的絕對(duì)路徑（按文件名排序）
    """
    if not os.path.exists(pdf_dir):
        raise FileNotFoundError(f"PDF目錄不存在：{pdf_dir}")
    
    # 匹配所有.pdf文件（不區(qū)分大小寫）
    pdf_pattern = os.path.join(pdf_dir, "*.pdf")
    pdf_files = glob.glob(pdf_pattern, recursive=False)
    
    # 按文件名排序，確保處理順序一致
    pdf_files.sort()
    
    if not pdf_files:
        print(f"警告：目錄 {pdf_dir} 中未找到PDF文件")
        return []
    
    print(f"成功找到 {len(pdf_files)} 個(gè)PDF文件：")
    for idx, pdf_file in enumerate(pdf_files, start=1):
        print(f"  {idx}. {os.path.basename(pdf_file)}")
    return pdf_files

from pdf2image import convert_from_path
from pathlib import Path

def pdf_to_png(
    pdf_path: str,
    root_image_dir: str = "pdf_images",
    dpi: int = 300,
    fmt: str = "png"
) -> List[str]:
    """
    PDF按頁(yè)轉(zhuǎn)PNG，每個(gè)PDF生成獨(dú)立目錄，返回圖片路徑列表（按頁(yè)碼排序）
    """
    pdf_filename = os.path.splitext(os.path.basename(pdf_path))[0]
    pdf_image_dir = os.path.join(root_image_dir, pdf_filename)
    
    # 創(chuàng)建獨(dú)立目錄（支持多級(jí)目錄自動(dòng)創(chuàng)建）
    Path(pdf_image_dir).mkdir(parents=True, exist_ok=True)
    print(f"  圖片輸出目錄：{os.path.abspath(pdf_image_dir)}")
    
    # Mac/Linux可自動(dòng)識(shí)別Poppler，Windows需指定路徑
    poppler_path = "/opt/homebrew/Cellar/poppler/25.12.0/bin" if os.name != "nt" else None
    if os.name == "nt" and not poppler_path:
        raise ValueError("Windows系統(tǒng)需指定Poppler的bin目錄路徑")
    
    try:
        # 按頁(yè)轉(zhuǎn)換PDF為圖片（保持原順序）
        images = convert_from_path(
            pdf_path=pdf_path,
            dpi=dpi,  # 分辨率越高識(shí)別越準(zhǔn)，建議≥200
            thread_count=4,  # 多線程提速
            fmt=fmt,
            grayscale=False,
            poppler_path=poppler_path
        )
        
        image_paths = []
        for page_num, image in enumerate(images, start=1):
            output_filename = f"{pdf_filename}_page_{page_num:02d}.{fmt}"
            output_path = os.path.join(pdf_image_dir, output_filename)
            image.save(output_path, fmt.upper())
            image_paths.append(output_path)
        
        print(f"  PDF轉(zhuǎn)PNG完成：共生成 {len(image_paths)} 張圖片")
        return image_paths
    except Exception as e:
        print(f"  PDF轉(zhuǎn)PNG失?。簕str(e)}")
        raise

import requests
import time
from typing import Optional

# Dify全局配置（需替換為實(shí)際信息）
DIFY_API_KEY = "app-xxxxxxxx"
DIFY_API_BASE_URL = "https://api.dify.ai/v1"
DIFY_USER = "difyuser"
RETRY_TIMES = 3
RETRY_DELAY = 2

def upload_image_to_dify(image_path: str) -> Optional[str]:
    """上傳圖片到Dify，返回文件ID"""
    upload_url = f"{DIFY_API_BASE_URL}/files/upload"
    headers = {"Authorization": f"Bearer {DIFY_API_KEY}"}
    
    try:
        with open(image_path, 'rb') as file:
            files = {'file': (os.path.basename(image_path), file, 'image/png')}
            data = {"user": DIFY_USER, "type": "IMAGE"}
            
            response = requests.post(upload_url, headers=headers, files=files, data=data)
            if response.status_code == 201:
                file_id = response.json().get("id")
                print(f"    圖片上傳成功，file_id：{file_id[:10]}...")
                return file_id
            else:
                print(f"    圖片上傳失敗，狀態(tài)碼：{response.status_code}，響應(yīng)：{response.text}")
                return None
    except Exception as e:
        print(f"    圖片上傳異常：{str(e)}")
        return None

def run_ocr_workflow(file_id: str) -> Optional[str]:
    """調(diào)用Dify OCR工作流，返回識(shí)別文本"""
    workflow_url = f"{DIFY_API_BASE_URL}/workflows/run"
    headers = {
        "Authorization": f"Bearer {DIFY_API_KEY}",
        "Content-Type": "application/json"
    }
    
    data = {
        "inputs": {
            "photo": {
                "transfer_method": "local_file",
                "upload_file_id": file_id,
                "type": "image"
            }
        },
        "response_mode": "blocking",
        "user": DIFY_USER
    }
    
    try:
        response = requests.post(workflow_url, headers=headers, json=data)
        if response.status_code == 200:
            result = response.json()
            # 解析Dify響應(yīng)結(jié)構(gòu)（data→outputs→text）
            if "data" in result and result["data"]["status"] == "succeeded":
                outputs = result["data"].get("outputs", {})
                ocr_text = outputs.get("text", "").strip()
                if ocr_text:
                    print(f"    OCR識(shí)別成功，文本長(zhǎng)度：{len(ocr_text)}字")
                    return ocr_text
            print(f"    OCR結(jié)果解析失敗：{result}")
            return None
        else:
            print(f"    工作流調(diào)用失敗，狀態(tài)碼：{response.status_code}，響應(yīng)：{response.text}")
            return None
    except Exception as e:
        print(f"    工作流調(diào)用異常：{str(e)}")
        return None

def batch_ocr_and_merge(
    image_paths: List[str],
    pdf_path: str,
    output_dir: str = "ocr_results",
    output_suffix: str = "txt"
) -> None:
    """批量OCR識(shí)別并按頁(yè)碼順序合并文本"""
    total_images = len(image_paths)
    merged_text = []
    
    # 創(chuàng)建文本輸出目錄
    Path(output_dir).mkdir(parents=True, exist_ok=True)
    pdf_filename = os.path.splitext(os.path.basename(pdf_path))[0]
    output_text_path = os.path.join(output_dir, f"{pdf_filename}.{output_suffix}")
    
    # 跳過(guò)已處理文件，避免重復(fù)勞動(dòng)
    if os.path.exists(output_text_path):
        print(f"  跳過(guò)已處理文件：{os.path.basename(output_text_path)}")
        return
    
    print(f"  開(kāi)始OCR處理（共{total_images}張圖片）...")
    for idx, image_path in enumerate(image_paths, start=1):
        print(f"\n    處理第{idx}/{total_images}張：{os.path.basename(image_path)}")
        
        # 圖片上傳（帶重試機(jī)制）
        file_id = None
        for retry in range(RETRY_TIMES):
            file_id = upload_image_to_dify(image_path)
            if file_id:
                break
            print(f"    第{retry+1}次上傳重試...")
            time.sleep(RETRY_DELAY)
        
        if not file_id:
            merged_text.extend([f"===== 第{idx}頁(yè)（圖片上傳失?。?=====", "\n"])
            continue
        
        # OCR識(shí)別（帶重試機(jī)制）
        ocr_text = None
        for retry in range(RETRY_TIMES):
            ocr_text = run_ocr_workflow(file_id)
            if ocr_text:
                break
            print(f"    第{retry+1}次OCR重試...")
            time.sleep(RETRY_DELAY)
        
        if ocr_text:
            merged_text.extend([f"===== 第{idx}頁(yè) =====", ocr_text, "\n"])
        else:
            merged_text.extend([f"===== 第{idx}頁(yè)（OCR識(shí)別失?。?=====", "\n"])
    
    # 保存合并文本（UTF-8編碼避免中文亂碼）
    final_text = "\n".join(merged_text)
    with open(output_text_path, "w", encoding="utf-8") as f:
        f.write(final_text)
    
    print(f"\n  文本合并完成！保存路徑：{os.path.abspath(output_text_path)}")
    print(f"  合并后總文本長(zhǎng)度：{len(final_text)}字")

def main():
    # 配置參數(shù)（可根據(jù)實(shí)際需求調(diào)整）
    PDF_DIR = "./pdf_files"  # PDF文件存放目錄
    ROOT_IMAGE_DIR = "pdf_images"  # 圖片存儲(chǔ)根目錄
    OCR_OUTPUT_DIR = "ocr_results"  # 文本輸出目錄
    OUTPUT_SUFFIX = "txt"  # 輸出文件后綴（支持txt/md）
    PDF_DPI = 300  # 圖片分辨率（建議200-300dpi）
    
    try:
        # 步驟1：獲取所有PDF文件
        pdf_files = get_all_pdf_in_dir(PDF_DIR)
        if not pdf_files:
            return
        
        # 步驟2：批量處理每個(gè)PDF
        total_pdfs = len(pdf_files)
        for pdf_idx, pdf_path in enumerate(pdf_files, start=1):
            print(f"\n=== 處理第{pdf_idx}/{total_pdfs}個(gè)PDF：{os.path.basename(pdf_path)} ===")
            try:
                # 子步驟1：PDF轉(zhuǎn)PNG
                image_paths = pdf_to_png(pdf_path, ROOT_IMAGE_DIR, PDF_DPI)
                # 子步驟2：OCR識(shí)別+文本合并
                batch_ocr_and_merge(image_paths, pdf_path, OCR_OUTPUT_DIR, OUTPUT_SUFFIX)
                print(f"=== 第{pdf_idx}/{total_pdfs}個(gè)PDF處理完成！ ===\n" + "-"*50)
            except Exception as e:
                print(f"=== 第{pdf_idx}/{total_pdfs}個(gè)PDF處理失?。簕str(e)} ===\n" + "-"*50)
                continue
        
        print(f"\n所有PDF文件處理完畢！")
        print(f"?? 圖片文件存放于：{os.path.abspath(ROOT_IMAGE_DIR)}")
        print(f"?? OCR文本文件存放于：{os.path.abspath(OCR_OUTPUT_DIR)}")
    except Exception as e:
        print(f"\n=== 批量處理流程異常終止：{str(e)} ===")

if __name__ == "__main__":
    main()