update embedding

app.py

@@ -50,12 +50,13 @@ def feedBack(context, story, chatbot=[], input=""):
 
 customerEmbedding = CustomEmbedding()
 
-faqChain = customerEmbedding.getFAQChain()
+faqChain = customerEmbedding.getFAQAgent()
 
 code_agent_executor = code_agent_executor()
 def faqFromLocal(input, chatbot=[]):
-    response = faqChain({"question": f"{input}"})
-    chatbot.append((input, response["answer"]))
+    # response = faqChain({"question": f"{input}"})
+    response = faqChain.run(input)
+    chatbot.append((input, response))
     return chatbot, ""
 
 

documents/abstract.faiss/index.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:65b241ca9d637fc607f43c0190c682677b635dbd36cddb0b754c0f74ea6988da
+size 26724

embedding.py

@@ -9,34 +9,35 @@ from langchain.chains.question_answering import load_qa_chain
 from langchain.document_loaders import NotionDirectoryLoader
 from langchain.memory import ConversationBufferMemory
 from langchain.chains import ConversationalRetrievalChain
+from langchain.agents import initialize_agent, AgentType, Tool, ZeroShotAgent, AgentExecutor
 
 from models import llm
 
 
 class CustomEmbedding:
     notionDirectoryLoader = NotionDirectoryLoader(
-        "documents/bussiness_context")
+        "/Users/peichao.dong/Documents/projects/dpc/ABstract/docs/pages")
     embeddings = HuggingFaceEmbeddings()
 
     def calculateEmbedding(self):
         documents = self.notionDirectoryLoader.load()
-        text_splitter = SpacyTextSplitter(
-            chunk_size=2048, pipeline="zh_core_web_sm", chunk_overlap=0)
+        # text_splitter = SpacyTextSplitter(
+        #     chunk_size=2048, pipeline="zh_core_web_sm", chunk_overlap=0)
         
-        # text_splitter = MarkdownTextSplitter(
-        #     chunk_size=4000, chunk_overlap=0)
+        text_splitter = MarkdownTextSplitter(
+            chunk_size=2048, chunk_overlap=0)
         texts = text_splitter.split_documents(documents)
 
         docsearch = FAISS.from_documents(texts, self.embeddings)
         docsearch.save_local(
-            folder_path="./documents/business_context.faiss")
+            folder_path="./documents/abstract.faiss")
 
     
 
     def getFAQChain(self, llm=llm(temperature=0.7)):
         memory = ConversationBufferMemory(memory_key="chat_history", return_messages=True)
         docsearch = FAISS.load_local(
-            "./documents/business_context.faiss", self.embeddings)
+            "./documents/abstract.faiss", self.embeddings)
         # retriever = VectorStoreRetriever(vectorstore=docsearch)
         _template = """Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question in chinese.
 
@@ -48,15 +49,45 @@ class CustomEmbedding:
         question_generator = LLMChain(llm=llm, prompt=CONDENSE_QUESTION_PROMPT)
         
         doc_chain = load_qa_chain(llm, chain_type="map_reduce")
-        qa = ConversationalRetrievalChain( retriever= docsearch.as_retriever(),
+        qa = ConversationalRetrievalChain( retriever= docsearch.as_retriever(search_kwargs={"k": 1}),
                                                     question_generator=question_generator,
                                                     combine_docs_chain=doc_chain,
                                                     memory=memory)
         return qa
 
+    def faq(self, input):
+        qa = self.getFAQChain()
+        response = qa({"question": f"{input}"})
+        return response["answer"]
 
-# customerEmbedding = CustomEmbedding()
-# # customerEmbedding.calculateEmbedding()
+    def getFAQAgent(self):
+        tools = [Tool(name="ABstract system FAQ", func= self.faq, description="Useful for anwer questions about ABstract system")]
+        memory = ConversationBufferMemory(memory_key="chat_history")
+
+        prefix = """Have a conversation with a human, answering the following questions as best you can. You have access to the following tools:"""
+        suffix = """Begin!"
+
+        {chat_history}
+        Question: {input}
+        {agent_scratchpad}"""
+
+        prompt = ZeroShotAgent.create_prompt(
+            tools, 
+            prefix=prefix, 
+            suffix=suffix, 
+            input_variables=["input", "chat_history", "agent_scratchpad"]
+        )
+
+        llm_chain = LLMChain(llm=llm(), prompt=prompt)
+        agent = ZeroShotAgent(llm_chain=llm_chain, tools=tools, verbose=True)
+        faq_agent = AgentExecutor.from_agent_and_tools(agent=agent, tools=tools, verbose=True, memory=memory)
+        return faq_agent
+        # faq_agent = initialize_agent(tools= tools, llm=llm(), agent=AgentType.CHAT_CONVERSATIONAL_REACT_DESCRIPTION, verbose=True)
+
+
+if __name__ == "__main__":
+    customerEmbedding = CustomEmbedding()
+    customerEmbedding.calculateEmbedding()
 # # customerEmbedding.calculateNotionEmbedding()
 
 # faq_chain = customerEmbedding.getFAQChain()