Update app.py

app.py

@@ -13,7 +13,16 @@ from langchain.vectorstores import Chroma
 import os
 from tqdm import tqdm
 import gradio as gr
-
+from langchain.memory import ConversationSummaryMemory
+from langchain.chains import ConversationalRetrievalChain
+from langchain.text_splitter import RecursiveCharacterTextSplitter
+from langchain.chains import LLMChain
+from langchain.prompts import (
+    ChatPromptTemplate,
+    HumanMessagePromptTemplate,
+    MessagesPlaceholder,
+    SystemMessagePromptTemplate,
+)
 documents=[]
 path='./bios/'
 # path='./augBios/'
@@ -24,15 +33,56 @@ for file in os.listdir(path):
     documents += loader.load()
 # print(documents)
 print(len(documents))
-text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=0)
+'''This is the code used for without memory chat'''
+# text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=0)
+# texts = text_splitter.split_documents(documents)
+
+# # embeddings = OpenAIEmbeddings()
+# embeddings = CohereEmbeddings(model='embed-english-v3.0')
+# docsearch = Chroma.from_documents(texts, embeddings)
+
+# qa = RetrievalQA.from_chain_type(llm=Cohere(model='command'), chain_type="stuff", \
+#     retriever=docsearch.as_retriever(search_kwargs={'k':1}),return_source_documents=True)
+
+# def predict(message, history):
+#     # history_langchain_format = []
+#     # for human, ai in history:
+#     #     history_langchain_format.append(HumanMessage(content=human))
+#     #     history_langchain_format.append(AIMessage(content=ai))
+#     # history_langchain_format.append(HumanMessage(content=message))
+#     # gpt_response = llm(history_langchain_format)
+#     # return gpt_response.content
+    
+#     message=message+'? just the book title-Author'
+#     #---------
+#     for human, ai in history:
+#         history_langchain_format.append(HumanMessage(content=human))
+#         history_langchain_format.append(AIMessage(content=ai))
+#     history_langchain_format.append(HumanMessage(content=message))
+#     gpt_response = llm(history_langchain_format)
+#     #------------
+#     return gpt_response.content
+#     result = qa({"query": message})
+#     # r1=docsearch.similarity_search_with_score(query=q,k=3)
+#     # print([(item[-2].metadata,item[-1]) for item in r1],\
+#     #       '\n\n',result['result'],f'|| {result["source_documents"][0].metadata}','\n*****\n')
+#     if result['result'] not in ["I don't know","I don't know."]:
+#         return result['result']+f'\n---\nAmazon Kindle ebook description is:\n {result["source_documents"][0].page_content}'+\
+#         f'\nfrom this file: {result["source_documents"][0].metadata}'
+#     else:
+#         return result['result']
+'''This is code used for with memory chat'''    
+text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=0)
 texts = text_splitter.split_documents(documents)
 
 # embeddings = OpenAIEmbeddings()
 embeddings = CohereEmbeddings(model='embed-english-v3.0')
 docsearch = Chroma.from_documents(texts, embeddings)
-
-qa = RetrievalQA.from_chain_type(llm=Cohere(model='command'), chain_type="stuff", \
-    retriever=docsearch.as_retriever(search_kwargs={'k':1}),return_source_documents=True)
+memory=ConversationSummaryMemory(
+    llm=Cohere(model='command'), memory_key="chat_history", return_messages=True
+)
+qa = ConversationalRetrievalChain.from_llm(llm=Cohere(model='command'), \
+    retriever=docsearch.as_retriever(),return_source_documents=True,memory=memory)
 
 def predict(message, history):
     # history_langchain_format = []
@@ -42,7 +92,17 @@ def predict(message, history):
     # history_langchain_format.append(HumanMessage(content=message))
     # gpt_response = llm(history_langchain_format)
     # return gpt_response.content
-    message=message+'? just the book title-Author'
+    
+    # message=message+'? just the book title-Autho'
+    #---------
+    # for human, ai in history:
+    #     history_langchain_format.append(HumanMessage(content=human))
+    #     history_langchain_format.append(AIMessage(content=ai))
+    # history_langchain_format.append(HumanMessage(content=message))
+    # gpt_response = llm(history_langchain_format)
+    # return gpt_response.content
+    #------------
+    
     result = qa({"query": message})
     # r1=docsearch.similarity_search_with_score(query=q,k=3)
     # print([(item[-2].metadata,item[-1]) for item in r1],\
@@ -51,10 +111,8 @@ def predict(message, history):
         return result['result']+f'\n---\nAmazon Kindle ebook description is:\n {result["source_documents"][0].page_content}'+\
         f'\nfrom this file: {result["source_documents"][0].metadata}'
     else:
-        return result['result']
-    
-    
-
+        return result['result']    
+'''------'''
 gr.ChatInterface(predict,
     chatbot=gr.Chatbot(height='auto'),
     textbox=gr.Textbox(placeholder="Recommend a book on someone who..."),