add openai agent

app.py

@@ -1,7 +1,7 @@
 # from typing import Any, Coroutine
 import openai
 import os
-from langchain.vectorstores import Chroma
+# from langchain.vectorstores import Chroma
 from langchain.embeddings.openai import OpenAIEmbeddings
 from langchain.text_splitter import CharacterTextSplitter
 from langchain.chat_models import AzureChatOpenAI
@@ -19,6 +19,14 @@ from langchain.python import PythonREPL
 from langchain.chains import LLMMathChain
 from langchain.memory import ConversationBufferMemory
 from langchain.agents import ZeroShotAgent, AgentExecutor
+from langchain.agents import OpenAIMultiFunctionsAgent
+from langchain.prompts import MessagesPlaceholder
+from langchain.schema.messages import (
+    AIMessage,
+    BaseMessage,
+    FunctionMessage,
+    SystemMessage,
+)
 from langchain import LLMChain
 import azure.cognitiveservices.speech as speechsdk
 import requests
@@ -51,6 +59,9 @@ from langchain.docstore.document import Document
 
 
 memory = ConversationBufferMemory(memory_key="chat_history")
+
+memory_openai = ConversationBufferMemory(memory_key="memory", return_messages=True)
+
 # Custom document loaders
 class MyElmLoader(UnstructuredEmailLoader):
     """Wrapper to fallback to text/plain when default does not work"""
@@ -413,9 +424,15 @@ prompt = ZeroShotAgent.create_prompt(
     input_variables=["input", "chat_history", "agent_scratchpad"]
 )
 
+prompt_openai = OpenAIMultiFunctionsAgent.create_prompt(
+    system_message = SystemMessage(
+            content="You are a helpful AI assistant."),
+    extra_prompt_messages = MessagesPlaceholder(variable_name="memory"),
+)
+
 input_variables=["input", "chat_history", "agent_scratchpad"]
 
-agent2 = initialize_agent(tools, llm, 
+agent_ZEROSHOT_REACT = initialize_agent(tools, llm, 
                         #  agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
                          agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION,
                          verbose = True,
@@ -436,8 +453,14 @@ agent2 = initialize_agent(tools, llm,
                         )
 
 llm_chain = LLMChain(llm=llm, prompt=prompt)
+
+llm_chain_openai = LLMChain(llm=llm, prompt=prompt_openai, verbose=True)
+
 agent_core = ZeroShotAgent(llm_chain=llm_chain, tools=tools, verbose=True)
-agent = AgentExecutor.from_agent_and_tools(
+
+agent_core_openai = OpenAIMultiFunctionsAgent(llm_chain=llm_chain_openai, tools=tools, verbose=True)
+
+agent_ZEROSHOT_AGENT = AgentExecutor.from_agent_and_tools(
     agent=agent_core, 
     tools=tools, 
     verbose=True, 
@@ -446,15 +469,28 @@ agent = AgentExecutor.from_agent_and_tools(
     max_iterations = int(os.getenv("max_iterations")),
     early_stopping_method="generate",
     )
-agent.max_execution_time = int(os.getenv("max_iterations"))
-agent.handle_parsing_errors = True
-agent.early_stopping_method = "generate"
+
+agent_OPENAI_MULTI = AgentExecutor.from_agent_and_tools(
+    agent=agent_core_openai, 
+    tools=tools, 
+    verbose=True, 
+    memory=memory_openai,
+    handle_parsing_errors = True,
+    max_iterations = int(os.getenv("max_iterations")),
+    early_stopping_method="generate",
+    )
+
+# agent.max_execution_time = int(os.getenv("max_iterations"))
+# agent.handle_parsing_errors = True
+# agent.early_stopping_method = "generate"
+
+agent = agent_ZEROSHOT_AGENT
 
 print(agent.agent.llm_chain.prompt.template)
 # print(agent.agent.llm_chain.prompt)
 
 global vectordb
-vectordb = Chroma(persist_directory='db', embedding_function=embeddings)
+# vectordb = Chroma(persist_directory='db', embedding_function=embeddings)
 global vectordb_p
 vectordb_p = Pinecone.from_existing_index(index_name, embeddings)
 
@@ -610,15 +646,18 @@ def QAQuery_p(question: str):
     source = res['source_documents']
     return response, source
 
-def CreatDb():
-    global vectordb
-    loader = DirectoryLoader('./documents', glob='**/*.txt')
-    documents = loader.load()
-    text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=200)
-    split_docs = text_splitter.split_documents(documents)
-    print(split_docs)
-    vectordb = Chroma.from_documents(split_docs, embeddings, persist_directory='db')
-    vectordb.persist()
+# def CreatDb():
+#     '''
+#     Funtion to creat chromadb DB based on with all docs
+#     '''
+#     global vectordb
+#     loader = DirectoryLoader('./documents', glob='**/*.txt')
+#     documents = loader.load()
+#     text_splitter = CharacterTextSplitter(chunk_size=500, chunk_overlap=200)
+#     split_docs = text_splitter.split_documents(documents)
+#     print(split_docs)
+#     vectordb = Chroma.from_documents(split_docs, embeddings, persist_directory='db')
+#     vectordb.persist()
 
 def QAQuery(question: str):
     global vectordb