Update app.py

app.py

@@ -1,23 +1,20 @@
 # https://qiita.com/nekoniii3/items/5acf764af65212d9f04f
 
 import gradio as gr
-import random
-import time
 
 import os
 
 from langchain_community.document_loaders import PyMuPDFLoader
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-# from langchain_community.chat_models import ChatOpenAI
 from langchain_openai import ChatOpenAI
 from langchain_community.vectorstores import Chroma
 from langchain.chains import RetrievalQA
-from langchain_community.embeddings import OpenAIEmbeddings
-
+# from langchain_openai import OpenAIEmbeddings
+from langchain_community.embeddings import HuggingFaceEmbeddings
 
 
 os.environ["TOKENIZERS_PARALLELISM"] = "false"
-os.environ["OPENAI_API_KEY"] = "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"
+os.environ["OPENAI_API_KEY"] = "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"
 
 file_name1 = 'ALV2_ALV3DTU操作マニュアルDTU-V3SET01.pdf'
 file_name2 = 'ALV3PCサーバ_ソフトウェア操作マニュアル_画像ファイル名付.pdf'
@@ -42,7 +39,8 @@ texts3 = text_splitter.split_documents(documents3)
 texts4 = text_splitter.split_documents(documents4)
 texts = texts1 + texts2 + texts3 + texts4
 
-embeddings = OpenAIEmbeddings(model="text-embedding-ada-002") 
+# embeddings = OpenAIEmbeddings(model="text-embedding-ada-002") 
+embeddings = HuggingFaceEmbeddings(model_name="oshizo/sbert-jsnli-luke-japanese-base-lite")
 vectordb = Chroma.from_documents(texts, embeddings)
 llm = ChatOpenAI(model_name="gpt-3.5-turbo-16k", temperature=0.05)
 
@@ -60,14 +58,45 @@ def save_image_filepath(filepath: str):
     shutil.copy(filepath, './filepath{}'.format(file_extension))
     pass
 
+import boto3
+s3 = boto3.client('s3',
+        aws_access_key_id="AKIA6ENMUHYQ7KWAEV7Q",
+        aws_secret_access_key="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",
+        region_name='ap-northeast-1'
+)
+
+
+# 画像のURL出力機能
+def get_public_url(bucket, target_object_path):
+    """
+    対象のS3ファイルのURLを取得する
+
+    Parameters
+    ----------
+    bucket: string
+        S3のバケット名
+    target_object_path: string
+        取得したいS3内のファイルパス
+
+    Returns
+    ----------
+    url: string
+        S3上のオブジェクトのURL
+    """
+    bucket_location = s3.get_bucket_location(Bucket=bucket)
+    return "https://s3-{0}.amazonaws.com/{1}/{2}".format(
+        bucket_location['LocationConstraint'],
+        bucket,
+        target_object_path)
+
+import fitz
+doc1 = fitz.open(file_name1)
+doc2 = fitz.open(file_name2)
+
+import math
+
 with gr.Blocks() as demo:
     chatbot = gr.Chatbot()
-    # with gr.Row():
-    #     with gr.Column():
-    #         image_input_filepath = gr.Image(type='filepath')
-            # image_button_filepath = gr.Button("filepath")
-
-    # image_button_filepath.click(save_image_filepath, inputs=image_input_filepath)
 
     msg = gr.Textbox()
 
@@ -76,23 +105,56 @@ with gr.Blocks() as demo:
         reply=reply2['result']
 
         for sd in reply2["source_documents"]:
-            page_content = str(sd.page_content)
+            # page_content = str(sd.page_content)
             source = str(sd.metadata["source"])
-            page = str(sd.metadata["page"]+1).zfill(3)
-            print("PDF：" + source)
-            print("ページ：" + page)
+            page = sd.metadata["page"]+1
+            page_num = str(page).zfill(3)
+            # print("PDF：" + source)
+            # print("ページ：" + page_num)
         
-        reply = reply + '<a href="https://dcs.mediapress-net.com/iportal/cv.do?c=20958580000&pg=108&v=MIW10001&d=LINK_MIW">link</a>'
-
+            if source == file_name1:
+                # ページ画像のURLを取得
+                bucket='page.dtu.manual'
+                key='page'+page_num+'_raster.png'
+                url = get_public_url(bucket, key)
+                reply = reply + ' <a href='+url+'>'+page_num+'</a>'
+
+            elif source == file_name2:
+                # ページ画像のURLを取得
+                bucket='page.server.manual'
+                key='page'+page_num+'_raster.png'
+                url = get_public_url(bucket, key)
+                reply = reply + ' <a href='+url+'>'+page_num+'</a>'
+
+                # PDFに貼り付けある画像のURLを取得
+                bucket='image.server.manual'
+                page2 = doc2[page]
+                page_annotations = page2.annots()
+                for annotation in page_annotations:
+                    annotation_num = str(annotation).zfill(3)
+                    # 注釈のプロパティを取得
+                    key = annotation.info.get('content', '')  # ノート注釈のテキストを取得
+                    url = get_public_url(bucket, key)
+                    reply = reply + ' <a href='+url+'>'+key+'</a>'
+            elif source == file_name3:
+                page2 = str(math.floor(1+float(page_num)/2))
+                url = "https://dcs.mediapress-net.com/iportal/cv.do?c=20958580000&pg="+page2+"&v=MIW10001&d=LINK_MIW"
+                reply = reply + ' <a href="'+url+'">'+page2+'</a>'
+            elif source == file_name4:
+                page2 = str(math.floor(1+(486+float(page_num))/2))
+                url = "https://dcs.mediapress-net.com/iportal/cv.do?c=20958580000&pg="+page2+"&v=MIW10001&d=LINK_MIW"
+                reply = reply + ' <a href="'+url+'">'+page2+'</a>'
+            else:
+                exit(0)
+                    
         return "", history + [[user_message, reply]]
 
     def bot(history):
         yield history
-        # save_image_filepath("./IMG_yosuke2.jpg")
 
     msg.submit(user, [msg, chatbot], [msg, chatbot], queue=True).then(
         bot, chatbot, chatbot
     )
     
 demo.queue()
-demo.launch(share=True)
+demo.launch(auth=("root", "test123456"), share=True)