From b2bef1e9348fef4010ed713497fdea85fc751c66 Mon Sep 17 00:00:00 2001
From: YM <479443481@qq.com>
Date: 星期四, 08 五月 2025 16:55:31 +0800
Subject: [PATCH] 文档解析，获取章节结构，并获取图片转换为base64
---
.gitignore | 3
static/images/test.png | 0
vision_test.py | 157 ++++++++++++++++++++++++++++++++++++++++++++++++++++
static/doc/ZL格式(公开).docx | 0
4 files changed, 159 insertions(+), 1 deletions(-)
diff --git a/.gitignore b/.gitignore
index ddcc4f8..7bc84b4 100644
--- a/.gitignore
+++ b/.gitignore
@@ -4,4 +4,5 @@
/datas
/.conda
/docs
-/out*
\ No newline at end of file
+/out*
+__pycache__
\ No newline at end of file
diff --git "a/static/doc/ZL\346\240\274\345\274\217\050\345\205\254\345\274\200\051.docx" "b/static/doc/ZL\346\240\274\345\274\217\050\345\205\254\345\274\200\051.docx"
new file mode 100644
index 0000000..754aaee
--- /dev/null
+++ "b/static/doc/ZL\346\240\274\345\274\217\050\345\205\254\345\274\200\051.docx"
Binary files differ
diff --git a/static/images/test.png b/static/images/test.png
new file mode 100644
index 0000000..aca2fdd
--- /dev/null
+++ b/static/images/test.png
Binary files differ
diff --git a/vision_test.py b/vision_test.py
new file mode 100644
index 0000000..3e41c4c
--- /dev/null
+++ b/vision_test.py
@@ -0,0 +1,157 @@
+# -*- coding: utf-8 -*-
+#
+# @author: lyg
+# @date: 2025-5-7
+# @version: 1
+# @description:瑙嗚璇嗗埆鏂囨。鍐呭
+
+from langchain_openai.chat_models import ChatOpenAI
+from langchain_core.prompts import HumanMessagePromptTemplate, ChatPromptTemplate
+from langchain_core.messages import HumanMessage, SystemMessage
+from langchain_core.output_parsers import JsonOutputParser
+from docx import Document
+from PIL import Image
+from io import BytesIO
+import re
+import json
+import base64
+
+
+class VisionTest:
+ def __init__(self, file):
+ self.llm = ChatOpenAI(temperature=0, model="qwen2.5-72b-instruct",
+ base_url="https://dashscope.aliyuncs.com/compatible-mode/v1", api_key="sk-15ecf7e273ad4b729c7f7f42b542749e")
+ image = base64.b64encode(open(file, 'rb').read()).decode()
+ self.prompt = ChatPromptTemplate.from_messages([
+ SystemMessage("浣犳槸涓�涓祫娣辫蒋浠跺伐绋嬪笀锛岃鍒嗘瀽鍥剧墖鍥炵瓟闂銆�"),
+ HumanMessage(content=[
+ {"type": "text", "text": "{msg}"},
+ {
+ "type": "image_url",
+ "image_url": {"url": f"data:image/jpeg;base64,{image}"},
+ }
+ ])
+ ])
+
+ def run(self, msg):
+ chain = self.prompt | self.llm
+ resp = chain.invoke({"msg": msg})
+ print(resp.content)
+
+ # def get_document_chapters(doc_path):
+ # doc = Document(doc_path)
+ # chapters = []
+ # current_chapter = None
+
+ # for para in doc.paragraphs:
+ # if para.style.name.startswith('Heading'): # 妫�鏌ユ槸鍚︿负鏍囬鏍峰紡
+ # level = int(para.style.name.replace('Heading', '')) # 鑾峰彇鏍囬绾у埆
+ # current_chapter = {'level': level, 'title': para.text, 'content': []}
+ # chapters.append(current_chapter)
+ # elif current_chapter is not None:
+ # current_chapter['content'].append(para.text) # 娣诲姞鍐呭鍒板綋鍓嶇珷鑺�
+
+ # return chapters
+
+ def has_image(self, paragraph):
+ # 閫氳繃妫�鏌ML涓殑宓屽叆寮忓璞℃潵鍒ゆ柇鏄惁鏈夊浘鐗�
+ xml = paragraph._element.xml
+ return 'w:object' in xml or 'w:drawing' in xml
+
+ def convert_blob_to_png_base64(self, image_blob):
+ try:
+ # 鎵撳紑鍥剧墖
+ image = Image.open(BytesIO(image_blob))
+ # 鍒涘缓鍐呭瓨缂撳啿鍖�
+ buffer = BytesIO()
+ # 淇濆瓨涓篜NG鏍煎紡
+ image.save(buffer, format="PNG")
+ # 鑾峰彇PNG鏍煎紡鐨勪簩杩涘埗鏁版嵁
+ png_data = buffer.getvalue()
+ # 杞崲涓築ase64缂栫爜
+ base64_data = base64.b64encode(png_data).decode('utf-8')
+ return base64_data
+ except Exception as e:
+ print(f"Error: {e}")
+ return None
+
+ def get_image_blob(self, paragraph):
+ # 閬嶅巻娈佃惤涓殑鎵�鏈塕un瀵硅薄锛堝浘鐗囬�氬父鍦ㄥ崟鐙殑Run涓級
+ for run in paragraph.runs:
+ xml = run._element.xml
+ if xml.find('v:imagedata') != -1:
+ # 浣跨敤姝ｅ垯琛ㄨ揪寮忔煡鎵緍:id灞炴��
+ match = re.search(r'r:id="([^"]+)"', xml)
+ if match:
+ r_id = match.group(1)
+ if r_id:
+ # 鑾峰彇鍥剧墖淇℃伅
+ image_part = paragraph.part.rels[r_id].target_part
+ return image_part.blob
+ if xml.find('wp:inline') != -1 or xml.find('wp:anchor') != -1:
+ # 浣跨敤姝ｅ垯琛ㄨ揪寮忔煡鎵緍:embed灞炴��
+ match = re.search(r'r:embed="([^"]+)"', xml)
+ if match:
+ r_id = match.group(1)
+ if r_id:
+ # 鑾峰彇鍥剧墖淇℃伅
+ image_part = paragraph.part.rels[r_id].target_part
+ return image_part.blob
+ return None
+
+ def loadDoc(self):
+ doc = Document('./static/doc/ZL鏍煎紡(鍏紑).docx')
+ # 鎸夌収鏍囬鑾峰彇娈佃惤鐨勫眰绾х粨鏋�
+ titles = []
+ for paragraph in doc.paragraphs:
+ if paragraph.text != "":
+ # 鏂囧瓧涓嶄负绌�
+ if paragraph.style.base_style is not None:
+ # 鏈塨ase_style
+ if paragraph.style.base_style.name.startswith('Heading'):
+ # 鏄爣棰�
+ level = int(paragraph.style.base_style.name.split(' ')[-1])
+ obj = {}
+ obj["level"] = level
+ obj["text"] = paragraph.text
+ obj["child"] = []
+ titles.append(obj)
+ else:
+ length = len(titles)
+ if "child" in titles[length -1]:
+ obj = {}
+ obj["text"] = paragraph.text
+ titles[length -1]['child'].append(obj)
+ else:
+ # 娌℃湁base_style
+ length = len(titles)
+ obj = {}
+ obj["text"] = paragraph.text
+ if length > 0 and "child" in titles[length -1]:
+ # 濡傛灉鏄爣棰樺唴鐨刟ppend杩涙爣棰樼殑child
+ titles[length -1]['child'].append(obj)
+ else:
+ # 闈炴爣棰樺唴鐨勭洿鎺ユ斁鍦ㄧ涓�灞�
+ titles.append(obj)
+ else:
+ # 鏂囧瓧涓虹┖鏃讹紝鍙兘鏄浘鐗囨垨鑰呰〃鏍�
+ if self.has_image(paragraph):
+ # 褰撳墠娈佃惤涓哄浘鐗�
+ obj = {}
+ # 鑾峰彇鍥剧墖鐨刡lob
+ img = self.get_image_blob(paragraph)
+ if img is not None:
+ imgBase64 = self.convert_blob_to_png_base64(img)
+ if imgBase64 is not None:
+ obj["imgBase64"] = imgBase64
+ titles[length -1]['child'].append(obj)
+ # 鍦ㄨ繖閲屾墿灞曞垽鏂〃鏍�
+ print(titles)
+ # for para in doc.paragraphs:
+ # print(para.text)
+ # print('------------------------')
+
+if __name__ == '__main__':
+ vision = VisionTest("./images/test.png")
+ # vision.run("闂")
+ vision.loadDoc()
--
Gitblit v1.9.1