add image url post

README.md

@@ -22,7 +22,7 @@ conda虚拟环境：conda activate search_book
 准备工作：
 1、将习题图书拆分为整页整页的完整图片，并使用合合OCR解析整页图片，放入excel的page_ocr列。
 2、将习题图书拆分为一个个独立的题目图片，并使用合合OCR解析题目图片，将ocr结果放入question_ocr列。将题目图片和题目合合eOCR结果一起输入豆包embedding模型"doubao-embedding-vision-250615"得到一个1024维的向量，处理为json字符串后放入excel的“question_image_embedding”列里。
-3、后端将以上excel数据通过接口插入es数据库（excel文档样例：“12668836-2025.xlsx”），在原有拆书平台下载的excel文档中添加【question_ocr(题目的ocr结果)，source_image_url(整页图片url)，page_ocr(整页图片ocr)，question_image_embedding(题目ocr和题目原图的向量)】列
+3、后端将以上excel数据通过接口插入es数据库（excel文档样例：12668836-2025.xlsx）
 
 
 搜整页图片流程：
@@ -108,7 +108,6 @@ delete_es_book(book_id)
 # 3、搜整页图片接口
 
 def search_book_question(book_id, image_ocr):
-    # 3、搜书搜题
     url = 'http://localhost:31001/search_page'
     url = 'http://61.170.32.8:31001/search_page'
     headers = {
@@ -142,8 +141,6 @@ print(res)
 #### 4、搜题接口 （只用题目合合OCR结果）
 
 ```
-# 4、搜题（只用题目paddleOCR结果）接口
-
 def search_question_text(book_id, question_ocr):
     url = 'http://localhost:31001/search_question_text'
     # url = 'http://61.170.32.8:31001/search_question_text'
@@ -178,9 +175,7 @@ search_question_text(book_id, question_ocr)
 #### 5、搜题接口 （题目合合ocr和题目图片作为一个整体的向量）
 
 ```
-# 5、搜题接口 （题目合合ocr和题目图片作为一个整体的向量）
-
-def search_question_embedding(book_id:str, question_embedding):
+def search_question_embedding(book_id:str, question_embedding:str):
     url = 'http://localhost:31001/search_question_embedding'
     # url = 'http://61.170.32.8:31001/search_question_embedding'
     headers = {
@@ -190,7 +185,7 @@ def search_question_embedding(book_id:str, question_embedding):
     # 表单数据
     data = {
         "book_id": book_id,
-        "question_embedding": question_embedding,
+        "question_embedding": question_embedding, # 1024维
         "top_k": 1
     }
     data = json.dumps(data, ensure_ascii=False)
@@ -214,6 +209,42 @@ res=search_question_embedding(book_id, question_embedding)
 
 
 
+#### 6、搜题接口 （输入题目图片合合ocr和题目图片url链接）
+
+```
+def search_question_embedding2(book_id, question_image_url, question_image_ocr):
+    url = 'http://localhost:31001/search_question_embedding'
+    url = 'http://61.170.32.8:31001/search_question_embedding'
+    headers = {
+        'accept': 'application/json',
+    }
+
+    # 表单数据
+    data = {
+        "book_id": book_id,
+        "question_image_url": question_image_url,
+        "question_image_ocr": question_image_ocr,
+        "top_k": 1
+    }
+    data = json.dumps(data, ensure_ascii=False)
+    response = requests.post(url, headers=headers, data=data )
+    # 输出响应结果
+    print(f"状态码: {response.status_code}")
+    res = response.json()
+
+    bankId = res['es_search'][0]['_source']['bankId']
+    print("bankId: ", bankId)
+    res = json.dumps(res, ensure_ascii=False, indent=4)
+    return res
+    
+book_id = "12670279"
+question_image_ocr = "题目图片的合合ocr结果"
+question_image_url = "https://example.123.jpg"
+res=search_question_embedding(book_id, question_image_url, question_image_ocr)
+```
+
+
+
 #### 6、启动服务
 
 ```linux

api_service.py

@@ -5,7 +5,7 @@ from pathlib import Path
 from pydantic import BaseModel
 import uvicorn
 import logging
-
+from tools import doubao_image_embedding
 from save_es_database import EsHelper
 from prepare_data import prepareData
 
@@ -51,7 +51,9 @@ class SearchQuestionText(BaseModel):
 # 搜题目请求数据结构（仅输入题目图片的向量）
 class SearchQuestionEmbedding(BaseModel):
     book_id: str
-    question_embedding: str # 题目的图片的embedding ; json.dumps([])
+    question_image_url: str = None # 题目图片的url链接
+    question_image_ocr: str = None # 题目图片的合合ocr结果
+    question_embedding: str = None # 题目图片和合合ocr文本的豆包1024维向量
     top_k: int = 1
 
 # 搜题目请求数据结构（输入题目图片的向量 和题目图片的OCR结果）
@@ -225,37 +227,35 @@ async def search_question_text(input: SearchQuestionText):
 # 4、搜图中的某个题目（只使用题目图片和ocr的向量）
 @app.post("/search_question_embedding")
 async def search_question_embedding(input: SearchQuestionEmbedding):
-    logger_es.info(f"search_question_embedding : book_id: {input.book_id}")
+    logger_es.info(f"search_question_embedding :  {input}")
     book_id = input.book_id
-    question_embedding = input.question_embedding
+    question_image_url = input.question_image_url
+    question_image_ocr = input.question_image_ocr
     top_k = input.top_k
-
-    try:
-        question_embedding = json.loads(question_embedding)
-        res = await es.search_question_embedding(index_name, book_id, question_embedding, top_k)
-    except Exception as e:
-        logger_es.info(f"search_question_embedding : 搜索 book_id:{book_id}，出错：{e}。")
-        raise HTTPException(status_code=500, detail=f"search_question_embedding 接口 book_id:{book_id}，出错：{e}。")
-
-    return {"status": "success", "es_search": res}
-
-
-# 5、搜图中某个题目（联合文本搜索和向量搜索）
-@app.post("/search_question_text_and_embedding")
-async def search_question_text_and_embedding(input: SearchQuestionTextAndEmbedding):
-    logger_es.info(f"search_question_text_and_embedding : {input.book_id}")
-    book_id = input.book_id
-    question_ocr = input.question_ocr
     question_embedding = input.question_embedding
 
-    top_k = input.top_k
+    if not question_embedding and question_image_url and question_image_ocr:
+        try:
+            question_embedding = doubao_image_embedding(question_image_url, question_image_ocr)  # 豆包大模型做图片和文本的向量1024维向量
+        except Exception as e:
+            logger_es.info(f"search_question_embedding : {input}，豆包embedding error ：{e}。")
+            raise HTTPException(status_code=510, detail=f"search_question_embedding： 豆包embedding error: {e}。")
 
-    try:
-        question_embedding = json.loads(question_embedding)
-        res = await es.search_question_textAndEmbedding(index_name, book_id, question_ocr, question_embedding, top_k)
-    except Exception as e:
-        logger_es.info(f"search_question_text_and_embedding : 搜索 book_id:{book_id}，出错：{e}。")
-        raise HTTPException(status_code=500, detail=f"search_question_text_and_embedding 接口 book_id:{book_id}，出错：{e}。")
+        try:
+            res = await es.search_question_embedding(index_name, book_id, question_embedding, top_k)
+        except Exception as e:
+            logger_es.info(f"search_question_embedding : 搜索 book_id:{book_id}，出错：{e}。")
+            raise HTTPException(status_code=500, detail=f"search_question_embedding 接口 book_id:{book_id}，出错：{e}。")
+
+    elif question_embedding:
+        try:
+            question_embedding = json.loads(question_embedding)
+            res = await es.search_question_embedding(index_name, book_id, question_embedding, top_k)
+        except Exception as e:
+            logger_es.info(f"search_question_embedding : 搜索 book_id:{book_id}，出错：{e}。")
+            raise HTTPException(status_code=500, detail=f"search_question_embedding 接口 book_id:{book_id}，出错：{e}。")
+    else:
+        raise HTTPException(status_code=520, detail=f"the question_image_url must be efficient")
 
     return {"status": "success", "es_search": res}