feat: mineru_web (#555)

projects/README.md

@@ -3,3 +3,5 @@
 ## 项目列表
 
 - [llama_index_rag](./llama_index_rag/README.md): 基于 llama_index 构建轻量级 RAG 系统
+
+- [web_api](./web_api/README.md): PDF解析的restful api服务

projects/web_api/README.md

+## 安装
+
+MinerU
+
+```bash
+# mineru已安装则跳过此步骤
+
+git clone https://github.com/opendatalab/MinerU.git
+cd MinerU
+
+conda create -n MinerU python=3.10
+conda activate MinerU
+pip install .[full] --extra-index-url https://wheels.myhloli.com
+```
+
+第三方软件
+
+```bash
+cd projects/web_api
+pip install poetry
+portey install
+```
+
+接口文档
+```
+在浏览器打开 mineru-web接口文档.html
+```

projects/web_api/pyproject.toml

+[tool.poetry]
+name = "web-api"
+version = "0.1.0"
+description = ""
+authors = ["houlinfeng <m15237195947@163.com>"]
+readme = "README.md"
+
+[tool.poetry.dependencies]
+python = "^3.10"
+flask = "^3.0.3"
+flask-restful = "^0.3.10"
+flask-cors = "^5.0.0"
+flask-sqlalchemy = "^3.1.1"
+flask-migrate = "^4.0.7"
+flask-jwt-extended = "^4.6.0"
+flask-marshmallow = "^1.2.1"
+pyyaml = "^6.0.2"
+loguru = "^0.7.2"
+marshmallow-sqlalchemy = "^1.1.0"
+
+
+[build-system]
+requires = ["poetry-core"]
+build-backend = "poetry.core.masonry.api"

projects/web_api/web_api/__init__.py

+__all__ = ["common", "api"]
\ No newline at end of file

projects/web_api/web_api/api/__init__.py

+import os
+from .extentions import app, db, migrate, jwt, ma
+from common.web_hook import before_request
+from common.logger import setup_log
+
+root_dir = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))
+print("root_dir", root_dir)
+
+def _register_db(flask_app):
+    from common import import_models
+    db.init_app(flask_app)
+    with app.app_context():
+        db.create_all()
+
+
+def create_app(config):
+    """
+    Create and configure an instance of the Flask application
+    :param config:
+    :return:
+    """
+    app.static_folder = os.path.join(root_dir, "static")
+    if config is None:
+        config = {}
+    app.config.update(config)
+    setup_log(config)
+    _register_db(app)
+    migrate.init_app(app=app, db=db)
+    jwt.init_app(app=app)
+    ma.init_app(app=app)
+    from .analysis import analysis_blue
+    app.register_blueprint(analysis_blue)
+
+    app.before_request(before_request)
+
+    return app

projects/web_api/web_api/api/analysis/__init__.py

+from flask import Blueprint
+from ..extentions import Api
+from .upload_view import UploadPdfView
+from .analysis_view import AnalysisTaskView, AnalysisTaskProgressView
+from .img_md_view import ImgView, MdView
+from .task_view import TaskView, HistoricalTasksView, DeleteTaskView
+
+analysis_blue = Blueprint('analysis', __name__)
+
+api_v2 = Api(analysis_blue, prefix='/api/v2')
+api_v2.add_resource(UploadPdfView, '/analysis/upload_pdf')
+api_v2.add_resource(AnalysisTaskView, '/extract/task/submit')
+api_v2.add_resource(AnalysisTaskProgressView, '/extract/task/progress')
+api_v2.add_resource(ImgView, '/analysis/pdf_img')
+api_v2.add_resource(MdView, '/analysis/pdf_md')
+api_v2.add_resource(TaskView, '/extract/taskQueue')
+api_v2.add_resource(HistoricalTasksView, '/extract/list')
+api_v2.add_resource(DeleteTaskView, '/extract/task')
\ No newline at end of file

projects/web_api/web_api/api/analysis/analysis_view.py

+import json
+import threading
+from pathlib import Path
+from flask import request, current_app, url_for
+from flask_restful import Resource
+from .ext import find_file, task_state_map
+# from .formula_ext import formula_detection, formula_recognition
+from .serialization import AnalysisViewSchema
+from marshmallow import ValidationError
+from ..extentions import db
+from .models import AnalysisTask, AnalysisPdf
+from .pdf_ext import analysis_pdf_task
+from common.custom_response import generate_response
+
+
+class AnalysisTaskProgressView(Resource):
+
+    def get(self):
+        """
+        获取任务进度
+        :return:
+        """
+        params = request.args
+        id = params.get('id')
+        analysis_task = AnalysisTask.query.filter(AnalysisTask.id == id).first()
+        if not analysis_task:
+            return generate_response(code=400, msg="Invalid ID", msgZH="无效id")
+        match analysis_task.task_type:
+            case 'pdf':
+                analysis_pdf = AnalysisPdf.query.filter(AnalysisPdf.id == analysis_task.analysis_pdf_id).first()
+                file_url = url_for('analysis.uploadpdfview', filename=analysis_task.file_name, as_attachment=False)
+                if analysis_task.status == 0:
+                    data = {
+                        "state": task_state_map.get(analysis_task.status),
+                        "status": analysis_pdf.status,
+                        "url": file_url,
+                        "fileName": analysis_task.file_name,
+                        "content": [],
+                        "markdownUrl": [],
+                        "fullMdLink": "",
+                        "type": analysis_task.task_type,
+                    }
+                    return generate_response(data=data)
+                elif analysis_task.status == 1:
+                    if analysis_pdf.status == 1:  # 任务正常完成
+                        bbox_info = json.loads(analysis_pdf.bbox_info)
+                        md_link_list = json.loads(analysis_pdf.md_link_list)
+                        full_md_link = analysis_pdf.full_md_link
+                        data = {
+                            "state": task_state_map.get(analysis_task.status),
+                            "status": analysis_pdf.status,
+                            "url": file_url,
+                            "fileName": analysis_task.file_name,
+                            "content": bbox_info,
+                            "markdownUrl": md_link_list,
+                            "fullMdLink": full_md_link,
+                            "type": analysis_task.task_type,
+                        }
+                        return generate_response(data=data)
+                    else:  # 任务异常结束
+                        data = {
+                            "state": task_state_map.get(analysis_task.status),
+                            "status": analysis_pdf.status,
+                            "url": file_url,
+                            "fileName": analysis_task.file_name,
+                            "content": [],
+                            "markdownUrl": [],
+                            "fullMdLink": "",
+                            "type": analysis_task.task_type,
+                        }
+                        return generate_response(code=-60004, data=data, msg="Failed to retrieve PDF parsing progress",
+                                                 msgZh="无法获取PDF解析进度")
+                else:
+                    data = {
+                        "state": task_state_map.get(analysis_task.status),
+                        "status": analysis_pdf.status,
+                        "url": file_url,
+                        "fileName": analysis_task.file_name,
+                        "content": [],
+                        "markdownUrl": [],
+                        "fullMdLink": "",
+                        "type": analysis_task.task_type,
+                    }
+                    return generate_response(data=data)
+            case 'formula-detect':
+                pass
+            case 'formula-extract':
+                pass
+            case 'table-recogn':
+                return generate_response(code=400, msg="Not yet supported", msgZH="尚不支持")
+            case _:
+                return generate_response()
+
+
+class AnalysisTaskView(Resource):
+
+    def post(self):
+        """
+        提交任务
+        :return:
+        """
+        analysis_view_schema = AnalysisViewSchema()
+        try:
+            params = analysis_view_schema.load(request.get_json())
+        except ValidationError as err:
+            return generate_response(code=400, msg=err.messages)
+        file_key = params.get("fileKey")
+        file_name = params.get("fileName")
+        task_type = params.get("taskType")
+        is_ocr = params.get("isOcr", False)
+
+        pdf_upload_folder = current_app.config['PDF_UPLOAD_FOLDER']
+        upload_dir = f"{current_app.static_folder}/{pdf_upload_folder}"
+        file_path = find_file(file_key, upload_dir)
+        match task_type:
+            case 'pdf':
+                if not file_path:
+                    return generate_response(code=400, msg="FileKey is invalid, no PDF file found",
+                                             msgZH="fileKey无效，未找到pdf文件")
+                analysis_task = AnalysisTask.query.filter(AnalysisTask.status.in_([0, 2])).first()
+                file_name = Path(file_path).name
+                with db.auto_commit():
+                    analysis_pdf_object = AnalysisPdf(
+                        file_name=file_name,
+                        file_path=file_path,
+                        status=3 if analysis_task else 0,
+                    )
+                    db.session.add(analysis_pdf_object)
+                    db.session.flush()
+                    analysis_pdf_id = analysis_pdf_object.id
+                with db.auto_commit():
+                    analysis_task_object = AnalysisTask(
+                        file_key=file_key,
+                        file_name=file_name,
+                        task_type=task_type,
+                        is_ocr=is_ocr,
+                        status=2 if analysis_task else 0,
+                        analysis_pdf_id=analysis_pdf_id
+                    )
+                    db.session.add(analysis_task_object)
+                    db.session.flush()
+                    analysis_task_id = analysis_task_object.id
+                if not analysis_task:  # 已有同类型任务在执行，请等待执行完成
+                    file_stem = Path(file_path).stem
+                    pdf_analysis_folder = current_app.config['PDF_ANALYSIS_FOLDER']
+                    pdf_dir = f"{current_app.static_folder}/{pdf_analysis_folder}/{file_stem}"
+                    image_dir = f"{pdf_dir}/images"
+                    t = threading.Thread(target=analysis_pdf_task,
+                                         args=(pdf_dir, image_dir, file_path, is_ocr, analysis_pdf_id))
+                    t.start()
+                # 生成文件的URL路径
+                file_url = url_for('analysis.uploadpdfview', filename=file_name, as_attachment=False)
+                data = {
+                    "url": file_url,
+                    "fileName": file_name,
+                    "id": analysis_task_id
+                }
+                return generate_response(data=data)
+            case 'formula-detect':
+                # if not file_path:
+                #     return generate_response(code=400, msg="FileKey is invalid, no image file found",
+                #                              msgZH="fileKey无效，未找到图片")
+                # return formula_detection(file_path, upload_dir)
+                return generate_response(code=400, msg="Not yet supported", msgZH="功能待开发")
+            case 'formula-extract':
+                # if not file_path:
+                #     return generate_response(code=400, msg="FileKey is invalid, no image file found",
+                #                              msgZH="fileKey无效，未找到图片")
+                # return formula_recognition(file_path, upload_dir)
+                return generate_response(code=400, msg="Not yet supported", msgZH="功能待开发")
+            case 'table-recogn':
+                return generate_response(code=400, msg="Not yet supported", msgZH="功能待开发")
+            case _:
+                return generate_response(code=400, msg="Not yet supported", msgZH="参数不支持")
+
+    def put(self):
+        """
+        重新发起任务
+        :return:
+        """
+        params = json.loads(request.data)
+        id = params.get('id')
+        analysis_task = AnalysisTask.query.filter(AnalysisTask.id == id).first()
+        match analysis_task.task_type:
+            case 'pdf':
+                task_r_p = AnalysisTask.query.filter(AnalysisTask.status.in_([0, 2])).first()
+                if task_r_p:
+                    with db.auto_commit():
+                        analysis_pdf_object = AnalysisPdf.query.filter_by(id=analysis_task.analysis_pdf_id).first()
+                        analysis_pdf_object.status = 3
+                        db.session.add(analysis_pdf_object)
+                    with db.auto_commit():
+                        analysis_task.status = 2
+                        db.session.add(analysis_task)
+                else:
+                    with db.auto_commit():
+                        analysis_pdf_object = AnalysisPdf.query.filter_by(id=analysis_task.analysis_pdf_id).first()
+                        analysis_pdf_object.status = 0
+                        db.session.add(analysis_pdf_object)
+                    with db.auto_commit():
+                        analysis_task.status = 0
+                        db.session.add(analysis_task)
+
+                    pdf_upload_folder = current_app.config['PDF_UPLOAD_FOLDER']
+                    upload_dir = f"{current_app.static_folder}/{pdf_upload_folder}"
+                    file_path = find_file(analysis_task.file_key, upload_dir)
+                    file_stem = Path(file_path).stem
+                    pdf_analysis_folder = current_app.config['PDF_ANALYSIS_FOLDER']
+                    pdf_dir = f"{current_app.static_folder}/{pdf_analysis_folder}/{file_stem}"
+                    image_dir = f"{pdf_dir}/images"
+                    t = threading.Thread(target=analysis_pdf_task,
+                                         args=(pdf_dir, image_dir, file_path, analysis_task.is_ocr,
+                                               analysis_task.analysis_pdf_id))
+                    t.start()
+
+                # 生成文件的URL路径
+                file_url = url_for('analysis.uploadpdfview', filename=analysis_task.file_name, as_attachment=False)
+                data = {
+                    "url": file_url,
+                    "fileName": analysis_task.file_name,
+                    "id": analysis_task.id
+                }
+                return generate_response(data=data)
+            case 'formula-detect':
+                return generate_response(code=400, msg="Not yet supported", msgZH="功能待开发")
+            case 'formula-extract':
+                return generate_response(code=400, msg="Not yet supported", msgZH="功能待开发")
+            case 'table-recogn':
+                return generate_response(code=400, msg="Not yet supported", msgZH="功能待开发")
+            case _:
+                return generate_response(code=400, msg="Not yet supported", msgZH="参数不支持")

projects/web_api/web_api/api/analysis/ext.py

+import os
+
+task_state_map = {
+    0: "running",
+    1: "finished",
+    2: "pending",
+}
+
+
+def find_file(file_key, file_dir):
+    """
+    查询文件
+    :param file_key:  文件哈希
+    :param file_dir:  文件目录
+    :return:
+    """
+    pdf_path = ""
+    for root, subDirs, files in os.walk(file_dir):
+        for fileName in files:
+            if fileName.startswith(file_key):
+                pdf_path = os.path.join(root, fileName)
+                break
+        if pdf_path:
+            break
+    return pdf_path

projects/web_api/web_api/api/analysis/formula_ext.py

+import os
+import pkgutil
+import numpy as np
+import yaml
+import argparse
+import cv2
+from pathlib import Path
+from ultralytics import YOLO
+from unimernet.common.config import Config
+import unimernet.tasks as tasks
+from unimernet.processors import load_processor
+from magic_pdf.libs.config_reader import get_local_models_dir, get_device
+from torchvision import transforms
+from magic_pdf.pre_proc.ocr_span_list_modify import remove_overlaps_low_confidence_spans, remove_overlaps_min_spans
+from PIL import Image
+from common.ext import singleton_func
+from common.custom_response import generate_response
+
+
+def mfd_model_init(weight):
+    mfd_model = YOLO(weight)
+    return mfd_model
+
+
+def mfr_model_init(weight_dir, cfg_path, _device_='cpu'):
+    args = argparse.Namespace(cfg_path=cfg_path, options=None)
+    cfg = Config(args)
+    cfg.config.model.pretrained = os.path.join(weight_dir, "pytorch_model.bin")
+    cfg.config.model.model_config.model_name = weight_dir
+    cfg.config.model.tokenizer_config.path = weight_dir
+    task = tasks.setup_task(cfg)
+    model = task.build_model(cfg)
+    model = model.to(_device_)
+    vis_processor = load_processor('formula_image_eval', cfg.config.datasets.formula_rec_eval.vis_processor.eval)
+    return model, vis_processor
+
+
+@singleton_func
+class CustomPEKModel:
+    def __init__(self):
+        # PDF-Extract-Kit/models
+        models_dir = get_local_models_dir()
+        self.device = get_device()
+        loader = pkgutil.get_loader("magic_pdf")
+        root_dir = Path(loader.path).parent
+        # model_config目录
+        model_config_dir = os.path.join(root_dir, 'resources', 'model_config')
+        # 构建 model_configs.yaml 文件的完整路径
+        config_path = os.path.join(model_config_dir, 'model_configs.yaml')
+        with open(config_path, "r", encoding='utf-8') as f:
+            configs = yaml.load(f, Loader=yaml.FullLoader)
+
+        # 初始化公式检测模型
+        self.mfd_model = mfd_model_init(str(os.path.join(models_dir, configs["weights"]["mfd"])))
+
+        # 初始化公式解析模型
+        mfr_weight_dir = str(os.path.join(models_dir, configs["weights"]["mfr"]))
+        mfr_cfg_path = str(os.path.join(model_config_dir, "UniMERNet", "demo.yaml"))
+        self.mfr_model, mfr_vis_processors = mfr_model_init(mfr_weight_dir, mfr_cfg_path, _device_=self.device)
+        self.mfr_transform = transforms.Compose([mfr_vis_processors, ])
+
+
+def get_all_spans(layout_dets) -> list:
+    def remove_duplicate_spans(spans):
+        new_spans = []
+        for span in spans:
+            if not any(span == existing_span for existing_span in new_spans):
+                new_spans.append(span)
+        return new_spans
+
+    all_spans = []
+    # allow_category_id_list = [3, 5, 13, 14, 15]
+    """当成span拼接的"""
+    #  3: 'image', # 图片
+    #  5: 'table',       # 表格
+    #  13: 'inline_equation',     # 行内公式
+    #  14: 'interline_equation',      # 行间公式
+    #  15: 'text',      # ocr识别文本
+    for layout_det in layout_dets:
+        if layout_det.get("bbox") is not None:
+            # 兼容直接输出bbox的模型数据,如paddle
+            x0, y0, x1, y1 = layout_det["bbox"]
+        else:
+            # 兼容直接输出poly的模型数据，如xxx
+            x0, y0, _, _, x1, y1, _, _ = layout_det["poly"]
+        bbox = [x0, y0, x1, y1]
+        layout_det["bbox"] = bbox
+        all_spans.append(layout_det)
+    return remove_duplicate_spans(all_spans)
+
+
+def formula_predict(mfd_model, image):
+    """
+    公式检测
+    :param mfd_model:
+    :param image:
+    :return:
+    """
+    latex_filling_list = []
+    # 公式检测
+    mfd_res = mfd_model.predict(image, imgsz=1888, conf=0.25, iou=0.45, verbose=True)[0]
+    for xyxy, conf, cla in zip(mfd_res.boxes.xyxy.cpu(), mfd_res.boxes.conf.cpu(), mfd_res.boxes.cls.cpu()):
+        xmin, ymin, xmax, ymax = [int(p.item()) for p in xyxy]
+        new_item = {
+            'category_id': 13 + int(cla.item()),
+            'poly': [xmin, ymin, xmax, ymin, xmax, ymax, xmin, ymax],
+            'score': round(float(conf.item()), 2),
+            'latex': '',
+        }
+        latex_filling_list.append(new_item)
+    return latex_filling_list
+
+
+def formula_detection(file_path, upload_dir):
+    """
+    公式检测
+    :param file_path:  文件路径
+    :param upload_dir:  上传文件夹
+    :return:
+    """
+    try:
+        image_open = Image.open(file_path)
+    except IOError:
+        return generate_response(code=400, msg="params is not valid", msgZh="参数类型不是图片，无效参数")
+
+    filename = Path(file_path).name
+
+    # 获取图片宽高
+    width, height = image_open.size
+    # 转换为RGB，忽略透明度通道
+    rgb_image = image_open.convert('RGB')
+    # 保存转换后的图片
+    rgb_image.save(file_path)
+
+    # 初始化模型
+    cpm = CustomPEKModel()
+    # 初始化公式检测模型
+    mfd_model = cpm.mfd_model
+
+    image_conv = Image.open(file_path)
+    image_array = np.array(image_conv)
+    pdf_width = 1416
+    pdf_height = 1888
+
+    # 重置图片大小
+    scale = min(pdf_width // 2 / width, pdf_height // 2 / height)  # 缩放比例
+    nw = int(width * scale)
+    nh = int(height * scale)
+    image_resize = cv2.resize(image_array, (nw, nh), interpolation=cv2.INTER_LINEAR)
+    resize_image_path = f"{upload_dir}/resize_{filename}"
+    cv2.imwrite(resize_image_path, image_resize)
+    # 将重置的图片贴到pdf白纸中
+    x = (pdf_width - nw) // 2
+    y = (pdf_height - nh) // 2
+    new_img = Image.new('RGB', (pdf_width, pdf_height), 'white')
+    image_scale = Image.open(resize_image_path)
+    new_img.paste(image_scale, (x, y))
+
+    # 公式检测
+    latex_filling_list = formula_predict(mfd_model, new_img)
+
+    os.remove(resize_image_path)
+
+    # 将缩放图公式检测的坐标还原为原图公式检测的坐标
+    for item in latex_filling_list:
+        item_poly = item["poly"]
+        item["poly"] = [
+            (item_poly[0] - x) / scale,
+            (item_poly[1] - y) / scale,
+            (item_poly[2] - x) / scale,
+            (item_poly[3] - y) / scale,
+            (item_poly[4] - x) / scale,
+            (item_poly[5] - y) / scale,
+            (item_poly[6] - x) / scale,
+            (item_poly[7] - y) / scale,
+        ]
+
+    if not latex_filling_list:
+        return generate_response(code=1001, msg="detection fail", msgZh="公式检测失败，图片过小，无法检测")
+
+    spans = get_all_spans(latex_filling_list)
+    '''删除重叠spans中置信度较低的那些'''
+    spans, dropped_spans_by_confidence = remove_overlaps_low_confidence_spans(spans)
+    '''删除重叠spans中较小的那些'''
+    spans, dropped_spans_by_span_overlap = remove_overlaps_min_spans(spans)
+
+    return generate_response(data={
+        'layout': spans,
+    })
+
+
+def formula_recognition(file_path, upload_dir):
+    """
+    公式识别
+    :param file_path:  文件路径
+    :param upload_dir:  上传文件夹
+    :return:
+    """
+    try:
+        image_open = Image.open(file_path)
+    except IOError:
+        return generate_response(code=400, msg="params is not valid", msgZh="参数类型不是图片，无效参数")
+
+    filename = Path(file_path).name
+
+    # 获取图片宽高
+    width, height = image_open.size
+    # 转换为RGB，忽略透明度通道
+    rgb_image = image_open.convert('RGB')
+    # 保存转换后的图片
+    rgb_image.save(file_path)
+
+    image_conv = Image.open(file_path)
+    image_array = np.array(image_conv)
+    pdf_width = 1416
+    pdf_height = 1888
+
+    # 重置图片大小
+    scale = min(pdf_width // 2 / width, pdf_height // 2 / height)  # 缩放比例
+    nw = int(width * scale)
+    nh = int(height * scale)
+    image_resize = cv2.resize(image_array, (nw, nh), interpolation=cv2.INTER_LINEAR)
+    resize_image_path = f"{upload_dir}/resize_{filename}"
+    cv2.imwrite(resize_image_path, image_resize)
+    # 将重置的图片贴到pdf白纸中
+    x = (pdf_width - nw) // 2
+    y = (pdf_height - nh) // 2
+    new_img = Image.new('RGB', (pdf_width, pdf_height), 'white')
+    image_scale = Image.open(resize_image_path)
+    new_img.paste(image_scale, (x, y))
+    new_img_array = np.array(new_img)
+
+    # 初始化模型
+    cpm = CustomPEKModel()
+    # device
+    device = cpm.device
+    # 初始化公式检测模型
+    mfd_model = cpm.mfd_model
+    # 初始化公式解析模型
+    mfr_model = cpm.mfr_model
+    mfr_transform = cpm.mfr_transform
+    # 公式识别
+    latex_filling_list, mfr_res = formula_recognition(mfd_model, new_img_array, mfr_transform, device, mfr_model,
+                                                      image_open)
+
+    os.remove(resize_image_path)
+
+    # 将缩放图公式检测的坐标还原为原图公式检测的坐标
+    for item in latex_filling_list:
+        item_poly = item["poly"]
+        item["poly"] = [
+            (item_poly[0] - x) / scale,
+            (item_poly[1] - y) / scale,
+            (item_poly[2] - x) / scale,
+            (item_poly[3] - y) / scale,
+            (item_poly[4] - x) / scale,
+            (item_poly[5] - y) / scale,
+            (item_poly[6] - x) / scale,
+            (item_poly[7] - y) / scale,
+        ]
+
+    spans = get_all_spans(latex_filling_list)
+    '''删除重叠spans中置信度较低的那些'''
+    spans, dropped_spans_by_confidence = remove_overlaps_low_confidence_spans(spans)
+    '''删除重叠spans中较小的那些'''
+    spans, dropped_spans_by_span_overlap = remove_overlaps_min_spans(spans)
+
+    if not latex_filling_list:
+        width, height = image_open.size
+        latex_filling_list.append({
+            'category_id': 14,
+            'poly': [0, 0, width, 0, width, height, 0, height],
+            'score': 1,
+            'latex': mfr_res[0] if mfr_res else "",
+        })
+
+    return generate_response(data={
+        'layout': spans if spans else latex_filling_list,
+        "mfr_res": mfr_res
+    })

projects/web_api/web_api/api/analysis/img_md_view.py

+from pathlib import Path
+from flask import request, current_app, send_from_directory
+from flask_restful import Resource
+
+
+class ImgView(Resource):
+    def get(self):
+        """
+        获取pdf解析的图片
+        :return:
+        """
+        params = request.args
+        pdf = params.get('pdf')
+        filename = params.get('filename')
+        as_attachment = params.get('as_attachment')
+        if str(as_attachment).lower() == "true":
+            as_attachment = True
+        else:
+            as_attachment = False
+        file_stem = Path(pdf).stem
+        pdf_analysis_folder = current_app.config['PDF_ANALYSIS_FOLDER']
+        pdf_dir = f"{current_app.static_folder}/{pdf_analysis_folder}/{file_stem}"
+        image_dir = f"{pdf_dir}/images"
+        response = send_from_directory(image_dir, filename, as_attachment=as_attachment)
+        return response
+
+
+class MdView(Resource):
+    def get(self):
+        """
+        获取pdf解析的markdown
+        :return:
+        """
+        params = request.args
+        pdf = params.get('pdf')
+        filename = params.get('filename')
+        as_attachment = params.get('as_attachment')
+        if str(as_attachment).lower() == "true":
+            as_attachment = True
+        else:
+            as_attachment = False
+        file_stem = Path(pdf).stem
+        pdf_analysis_folder = current_app.config['PDF_ANALYSIS_FOLDER']
+        pdf_dir = f"{current_app.static_folder}/{pdf_analysis_folder}/{file_stem}"
+        response = send_from_directory(pdf_dir, filename, as_attachment=as_attachment)
+        return response

projects/web_api/web_api/api/analysis/models.py

+from datetime import datetime
+from ..extentions import db
+
+
+class AnalysisTask(db.Model):
+    __tablename__ = 'analysis_task'
+    id = db.Column(db.Integer, primary_key=True, autoincrement=True)
+    file_key = db.Column(db.Text, comment="文件唯一哈希")
+    file_name = db.Column(db.Text, comment="文件名称")
+    task_type = db.Column(db.String(128), comment="任务类型")
+    is_ocr = db.Column(db.Boolean, default=False, comment="是否ocr")
+    status = db.Column(db.Integer, default=0, comment="状态")  # 0 running  1 finished  2 pending
+    analysis_pdf_id = db.Column(db.Integer, comment="analysis_pdf的id")
+    create_date = db.Column(db.DateTime(), nullable=False, default=datetime.now)
+    update_date = db.Column(db.DateTime(), nullable=False, default=datetime.now, onupdate=datetime.now)
+
+
+class AnalysisPdf(db.Model):
+    __tablename__ = 'analysis_pdf'
+    id = db.Column(db.Integer, primary_key=True, autoincrement=True)
+    file_name = db.Column(db.Text, comment="文件名称")
+    file_url = db.Column(db.Text, comment="文件原路径")
+    file_path = db.Column(db.Text, comment="文件路径")
+    status = db.Column(db.Integer, default=3, comment="状态")  # 0 转换中  1 已完成  2 转换失败 3 init
+    bbox_info = db.Column(db.Text, comment="坐标数据")
+    md_link_list = db.Column(db.Text, comment="markdown分页链接")
+    full_md_link = db.Column(db.Text, comment="markdown全文链接")
+    create_date = db.Column(db.DateTime(), nullable=False, default=datetime.now)
+    update_date = db.Column(db.DateTime(), nullable=False, default=datetime.now, onupdate=datetime.now)
\ No newline at end of file

projects/web_api/web_api/api/analysis/pdf_ext.py

+import json
+import re
+import traceback
+from pathlib import Path
+from flask import current_app, url_for
+from magic_pdf.rw.DiskReaderWriter import DiskReaderWriter
+from magic_pdf.pipe.UNIPipe import UNIPipe
+import magic_pdf.model as model_config
+from magic_pdf.libs.json_compressor import JsonCompressor
+from magic_pdf.dict2md.ocr_mkcontent import ocr_mk_mm_markdown_with_para_and_pagination
+from .ext import find_file
+from ..extentions import app, db
+from .models import AnalysisPdf, AnalysisTask
+from common.error_types import ApiException
+from loguru import logger
+
+model_config.__use_inside_model__ = True
+
+
+def analysis_pdf(image_dir, pdf_bytes, is_ocr=False):
+    try:
+        model_json = []  # model_json传空list使用内置模型解析
+        logger.info(f"is_ocr: {is_ocr}")
+        if not is_ocr:
+            jso_useful_key = {"_pdf_type": "", "model_list": model_json}
+            image_writer = DiskReaderWriter(image_dir)
+            pipe = UNIPipe(pdf_bytes, jso_useful_key, image_writer, is_debug=True)
+            pipe.pipe_classify()
+        else:
+            jso_useful_key = {"_pdf_type": "ocr", "model_list": model_json}
+            image_writer = DiskReaderWriter(image_dir)
+            pipe = UNIPipe(pdf_bytes, jso_useful_key, image_writer, is_debug=True)
+        """如果没有传入有效的模型数据，则使用内置model解析"""
+        if len(model_json) == 0:
+            if model_config.__use_inside_model__:
+                pipe.pipe_analyze()
+            else:
+                logger.error("need model list input")
+                exit(1)
+        pipe.pipe_parse()
+        pdf_mid_data = JsonCompressor.decompress_json(pipe.get_compress_pdf_mid_data())
+        pdf_info_list = pdf_mid_data["pdf_info"]
+        md_content = json.dumps(ocr_mk_mm_markdown_with_para_and_pagination(pdf_info_list, image_dir),
+                                ensure_ascii=False)
+        bbox_info = get_bbox_info(pdf_info_list)
+        return md_content, bbox_info
+    except Exception as e:
+        logger.error(traceback.format_exc())
+
+
+def get_bbox_info(data):
+    bbox_info = []
+    for page in data:
+        preproc_blocks = page.get("preproc_blocks", [])
+        discarded_blocks = page.get("discarded_blocks", [])
+        bbox_info.append({
+            "preproc_blocks": preproc_blocks,
+            "page_idx": page.get("page_idx"),
+            "page_size": page.get("page_size"),
+            "discarded_blocks": discarded_blocks,
+        })
+    return bbox_info
+
+
+def analysis_pdf_task(pdf_dir, image_dir, pdf_path, is_ocr, analysis_pdf_id):
+    """
+    解析pdf
+    :param pdf_dir:  pdf解析目录
+    :param image_dir:  图片目录
+    :param pdf_path:  pdf路径
+    :param is_ocr:  是否启用ocr
+    :param analysis_pdf_id:  pdf解析表id
+    :return:
+    """
+    try:
+        logger.info(f"start task: {pdf_path}")
+        logger.info(f"image_dir: {image_dir}")
+        if not Path(image_dir).exists():
+            Path(image_dir).mkdir(parents=True, exist_ok=True)
+        with open(pdf_path, 'rb') as file:
+            pdf_bytes = file.read()
+        md_content, bbox_info = analysis_pdf(image_dir, pdf_bytes, is_ocr)
+        img_list = Path(image_dir).glob('*') if Path(image_dir).exists() else []
+
+        pdf_name = Path(pdf_path).name
+        with app.app_context():
+            for img in img_list:
+                img_name = Path(img).name
+                regex = re.compile(fr'.*\((.*?{img_name})')
+                regex_result = regex.search(md_content)
+                img_url = url_for('analysis.imgview', filename=img_name, as_attachment=False)
+                md_content = md_content.replace(regex_result.group(1), f"{img_url}&pdf={pdf_name}")
+
+        full_md_content = ""
+        for item in json.loads(md_content):
+            full_md_content += item["md_content"] + "\n"
+
+        full_md_name = "full.md"
+        with open(f"{pdf_dir}/{full_md_name}", "w") as file:
+            file.write(full_md_content)
+        with app.app_context():
+            full_md_link = url_for('analysis.mdview', filename=full_md_name, as_attachment=False)
+            full_md_link = f"{full_md_link}&pdf={pdf_name}"
+
+        md_link_list = []
+        with app.app_context():
+            for n, md in enumerate(json.loads(md_content)):
+                md_content = md["md_content"]
+                md_name = f"{md.get('page_no', n)}.md"
+                with open(f"{pdf_dir}/{md_name}", "w") as file:
+                    file.write(md_content)
+                md_url = url_for('analysis.mdview', filename=md_name, as_attachment=False)
+                md_link_list.append(f"{md_url}&pdf={pdf_name}")
+
+        with app.app_context():
+            with db.auto_commit():
+                analysis_pdf_object = AnalysisPdf.query.filter_by(id=analysis_pdf_id).first()
+                analysis_pdf_object.status = 1
+                analysis_pdf_object.bbox_info = json.dumps(bbox_info, ensure_ascii=False)
+                analysis_pdf_object.md_link_list = json.dumps(md_link_list, ensure_ascii=False)
+                analysis_pdf_object.full_md_link = full_md_link
+                db.session.add(analysis_pdf_object)
+            with db.auto_commit():
+                analysis_task_object = AnalysisTask.query.filter_by(analysis_pdf_id=analysis_pdf_id).first()
+                analysis_task_object.status = 1
+                db.session.add(analysis_task_object)
+        logger.info(f"finished!")
+    except Exception as e:
+        logger.error(traceback.format_exc())
+        with app.app_context():
+            with db.auto_commit():
+                analysis_pdf_object = AnalysisPdf.query.filter_by(id=analysis_pdf_id).first()
+                analysis_pdf_object.status = 2
+                db.session.add(analysis_pdf_object)
+            with db.auto_commit():
+                analysis_task_object = AnalysisTask.query.filter_by(analysis_pdf_id=analysis_pdf_id).first()
+                analysis_task_object.status = 1
+                db.session.add(analysis_task_object)
+        raise ApiException(code=500, msg="PDF parsing failed", msgZH="pdf解析失败")
+    finally:
+        # 执行pending
+        with app.app_context():
+            analysis_task_object = AnalysisTask.query.filter_by(status=2).order_by(
+                AnalysisTask.update_date.asc()).first()
+            if analysis_task_object:
+                pdf_upload_folder = current_app.config['PDF_UPLOAD_FOLDER']
+                upload_dir = f"{current_app.static_folder}/{pdf_upload_folder}"
+                file_path = find_file(analysis_task_object.file_key, upload_dir)
+                file_stem = Path(file_path).stem
+                pdf_analysis_folder = current_app.config['PDF_ANALYSIS_FOLDER']
+                pdf_dir = f"{current_app.static_folder}/{pdf_analysis_folder}/{file_stem}"
+                image_dir = f"{pdf_dir}/images"
+                with db.auto_commit():
+                    analysis_pdf_object = AnalysisPdf.query.filter_by(id=analysis_task_object.analysis_pdf_id).first()
+                    analysis_pdf_object.status = 0
+                    db.session.add(analysis_pdf_object)
+                with db.auto_commit():
+                    analysis_task_object.status = 0
+                    db.session.add(analysis_task_object)
+                analysis_pdf_task(pdf_dir, image_dir, file_path, analysis_task_object.is_ocr, analysis_task_object.analysis_pdf_id)
+            else:
+                logger.info(f"all task finished!")

projects/web_api/web_api/api/analysis/serialization.py

+from marshmallow import Schema, fields, validates_schema, validates
+from common.error_types import ApiException
+from .models import AnalysisTask
+
+
+class BooleanField(fields.Boolean):
+    def _deserialize(self, value, attr, data, **kwargs):
+        # 进行自定义验证
+        if not isinstance(value, bool):
+            raise ApiException(code=400, msg="isOcr not a valid boolean", msgZH="isOcr不是有效的布尔值")
+
+        return value
+
+
+class AnalysisViewSchema(Schema):
+    fileKey = fields.Str(required=True)
+    fileName = fields.Str()
+    taskType = fields.Str(required=True)
+    isOcr = BooleanField()
+
+    @validates_schema(pass_many=True)
+    def validate_passwords(self, data, **kwargs):
+        task_type = data['taskType']
+        file_key = data['fileKey']
+        if not file_key:
+            raise ApiException(code=400, msg="fileKey cannot be empty", msgZH="fileKey不能为空")
+        if not task_type:
+            raise ApiException(code=400, msg="taskType cannot be empty", msgZH="taskType不能为空")

projects/web_api/web_api/api/analysis/task_view.py

+import json
+from flask import url_for, request
+from flask_restful import Resource
+from sqlalchemy import func
+from ..extentions import db
+from .models import AnalysisTask, AnalysisPdf
+from .ext import task_state_map
+from common.custom_response import generate_response
+
+
+class TaskView(Resource):
+    def get(self):
+        """
+        查询正在进行的任务
+        :return:
+        """
+        analysis_task_running = AnalysisTask.query.filter(AnalysisTask.status == 0).first()
+        analysis_task_pending = AnalysisTask.query.filter(AnalysisTask.status == 2).order_by(
+            AnalysisTask.create_date.asc()).all()
+        pending_total = db.session.query(func.count(AnalysisTask.id)).filter(AnalysisTask.status == 2).scalar()
+        task_nums = pending_total + 1
+        data = [
+            {
+                "queues": task_nums,  # 正在排队的任务总数
+                "rank": 1,
+                "id": analysis_task_running.id,
+                "url": url_for('analysis.uploadpdfview', filename=analysis_task_running.file_name, as_attachment=False),
+                "fileName": analysis_task_running.file_name,
+                "type": analysis_task_running.task_type,
+                "state": task_state_map.get(analysis_task_running.status),
+            }
+        ]
+        for n, task in enumerate(analysis_task_pending):
+            data.append({
+                "queues": task_nums,  # 正在排队的任务总数
+                "rank": n + 2,
+                "id": task.id,
+                "url": url_for('analysis.uploadpdfview', filename=task.file_name, as_attachment=False),
+                "fileName": task.file_name,
+                "type": task.task_type,
+                "state": task_state_map.get(task.status),
+            })
+        data.reverse()
+        return generate_response(data=data, total=task_nums)
+
+
+class HistoricalTasksView(Resource):
+    def get(self):
+        """
+        获取任务历史记录
+        :return:
+        """
+        params = request.args
+        page_no = params.get('pageNo', 1)
+        page_size = params.get('pageSize', 10)
+        total = db.session.query(func.count(AnalysisTask.id)).scalar()
+        analysis_task = AnalysisTask.query.order_by(AnalysisTask.create_date.desc()).paginate(page=int(page_no),
+                                                                                              per_page=int(page_size),
+                                                                                              error_out=False)
+        data = []
+        for n, task in enumerate(analysis_task):
+            data.append({
+                "fileName": task.file_name,
+                "id": task.id,
+                "type": task.task_type,
+                "state": task_state_map.get(task.status),
+            })
+        data = {
+            "list": data,
+            "total": total,
+            "pageNo": page_no,
+            "pageSize": page_size,
+        }
+        return generate_response(data=data)
+
+
+class DeleteTaskView(Resource):
+    def delete(self):
+        """
+        删除任务历史记录
+        :return:
+        """
+        params = json.loads(request.data)
+        id = params.get('id')
+
+        analysis_task = AnalysisTask.query.filter(AnalysisTask.id == id, AnalysisTask.status != 0).first()
+        if analysis_task:
+            analysis_pdf = AnalysisPdf.query.filter(AnalysisPdf.id == AnalysisTask.analysis_pdf_id).first()
+            with db.auto_commit():
+                db.session.delete(analysis_pdf)
+                db.session.delete(analysis_task)
+        else:
+            return generate_response(code=400, msg="The ID is incorrect", msgZH="id不正确")
+
+        return generate_response(data={"id": id})

projects/web_api/web_api/api/analysis/upload_view.py

+import json
+import traceback
+import requests
+from flask import request, current_app, url_for, send_from_directory
+from flask_restful import Resource
+from werkzeug.utils import secure_filename
+from pathlib import Path
+from common.ext import is_pdf, calculate_file_hash, url_is_pdf
+from io import BytesIO
+from werkzeug.datastructures import FileStorage
+from common.custom_response import generate_response
+from loguru import logger
+
+
+class UploadPdfView(Resource):
+
+    def get(self):
+        """
+        获取pdf
+        :return:
+        """
+        params = request.args
+        filename = params.get('filename')
+        as_attachment = params.get('as_attachment')
+        if str(as_attachment).lower() == "true":
+            as_attachment = True
+        else:
+            as_attachment = False
+        pdf_upload_folder = current_app.config['PDF_UPLOAD_FOLDER']
+        response = send_from_directory(f"{current_app.static_folder}/{pdf_upload_folder}", filename,
+                                       as_attachment=as_attachment)
+        return response
+
+    def post(self):
+        """
+        上传pdf
+        :return:
+        """
+        file_list = request.files.getlist("file")
+        if file_list:
+            file = file_list[0]
+            filename = secure_filename(file.filename)
+            if not file or file and not is_pdf(filename, file):
+                return generate_response(code=400, msg="Invalid PDF file", msgZH="PDF文件参数无效")
+        else:
+            params = json.loads(request.data)
+            pdf_url = params.get('pdfUrl')
+            try:
+                response = requests.get(pdf_url, stream=True)
+            except ConnectionError as e:
+                logger.error(traceback.format_exc())
+                return generate_response(code=400, msg="params is not valid", msgZh="参数错误，pdf链接无法访问")
+            if response.status_code != 200:
+                return generate_response(code=400, msg="params is not valid", msgZh="参数错误，pdf链接响应状态异常")
+            # 创建一个模拟的 FileStorage 对象
+            file_content = BytesIO(response.content)
+            filename = Path(pdf_url).name if ".pdf" in pdf_url else f"{Path(pdf_url).name}.pdf"
+            file = FileStorage(
+                stream=file_content,
+                filename=filename,
+                content_type=response.headers.get('Content-Type', 'application/octet-stream')
+            )
+            if not file or file and not url_is_pdf(file):
+                return generate_response(code=400, msg="Invalid PDF file", msgZH="PDF文件参数无效")
+
+        pdf_upload_folder = current_app.config['PDF_UPLOAD_FOLDER']
+        upload_dir = f"{current_app.static_folder}/{pdf_upload_folder}"
+        if not Path(upload_dir).exists():
+            Path(upload_dir).mkdir(parents=True, exist_ok=True)
+        file_key = calculate_file_hash(file)
+        # new_filename = f"{int(time.time())}_{filename}"
+        new_filename = f"{file_key}_{filename}"
+        file_path = f"{upload_dir}/{new_filename}"
+        # file.save(file_path)
+        chunk_size = 8192
+        with open(file_path, 'wb') as f:
+            while True:
+                chunk = file.stream.read(chunk_size)
+                if not chunk:
+                    break
+                f.write(chunk)
+
+        # 生成文件的URL路径
+        file_url = url_for('analysis.uploadpdfview', filename=new_filename, as_attachment=False)
+        data = {
+            "url": file_url,
+            "file_key": file_key
+        }
+        return generate_response(data=data)

projects/web_api/web_api/api/extentions.py

+from flask import Flask, jsonify
+from flask_restful import Api as _Api
+from flask_cors import CORS
+from flask_sqlalchemy import SQLAlchemy as _SQLAlchemy
+from flask_migrate import Migrate
+from contextlib import contextmanager
+from flask_jwt_extended import JWTManager
+from flask_marshmallow import Marshmallow
+from common.error_types import ApiException
+from werkzeug.exceptions import HTTPException
+from loguru import logger
+
+
+class Api(_Api):
+    def handle_error(self, e):
+        if isinstance(e, ApiException):
+            code = e.code
+            msg = e.msg
+            msgZH = e.msgZH
+            error_code = e.error_code
+        elif isinstance(e, HTTPException):
+            code = e.code
+            msg = e.description
+            msgZH = "服务异常，详细信息请查看日志"
+            error_code = e.code
+        else:
+            code = 500
+            msg = str(e)
+            error_code = 500
+            msgZH = "服务异常，详细信息请查看日志"
+
+        # 使用 loguru 记录异常信息
+        logger.opt(exception=e).error(f"An error occurred: {msg}")
+
+        return jsonify({
+            "error": "Internal Server Error" if code == 500 else e.name,
+            "msg": msg,
+            "msgZH": msgZH,
+            "code": code,
+            "error_code": error_code
+        }), code
+
+
+class SQLAlchemy(_SQLAlchemy):
+    @contextmanager
+    def auto_commit(self):
+        try:
+            yield
+            db.session.commit()
+            db.session.flush()
+        except Exception as e:
+            db.session.rollback()
+            raise e
+
+
+app = Flask(__name__)
+CORS(app, supports_credentials=True)
+db = SQLAlchemy()
+migrate = Migrate()
+jwt = JWTManager()
+ma = Marshmallow()

projects/web_api/web_api/app.py

+import socket
+from api import create_app
+from pathlib import Path
+import yaml
+
+
+def get_local_ip():
+    sock = socket.socket(socket.AF_INET, socket.SOCK_DGRAM)
+    sock.connect(('8.8.8.8', 80))  # Google DNS 服务器
+    ip_address = sock.getsockname()[0]
+    sock.close()
+    return ip_address
+
+
+current_file_path = Path(__file__).resolve()
+base_dir = current_file_path.parent
+config_path = base_dir / "config/config.yaml"
+
+
+class ConfigMap(dict):
+    __setattr__ = dict.__setitem__
+    __getattr__ = dict.__getitem__
+
+
+with open(str(config_path), mode='r', encoding='utf-8') as fd:
+    data = yaml.load(fd, Loader=yaml.FullLoader)
+    _config = data.get(data.get("CurrentConfig", "DevelopmentConfig"))
+config = ConfigMap()
+for k, v in _config.items():
+    config[k] = v
+config['base_dir'] = base_dir
+database = _config.get("database")
+if database:
+    if database.get("type") == "sqlite":
+        database_uri = f'sqlite:///{base_dir}/{database.get("path")}'
+    elif database.get("type") == "mysql":
+        database_uri = f'mysql+pymysql://{database.get("user")}:{database.get("password")}@{database.get("host")}:{database.get("port")}/{database.get("database")}?'
+    else:
+        database_uri = ''
+    config['SQLALCHEMY_DATABASE_URI'] = database_uri
+
+ip_address = get_local_ip()
+port = config.get("PORT", 5559)
+# 配置 SERVER_NAME
+config['SERVER_NAME'] = f'{ip_address}:5559'
+# 配置 APPLICATION_ROOT
+config['APPLICATION_ROOT'] = '/'
+# 配置 PREFERRED_URL_SCHEME
+config['PREFERRED_URL_SCHEME'] = 'http'
+
+app = create_app(config)
+
+if __name__ == '__main__':
+    app.run(host="0.0.0.0", port=port, debug=config.get("DEBUG", False))