app/simchat.py

import os
import time
import pinecone
import chainlit as cl

from promptwatch import PromptWatch
from prompts import load_query_gen_prompt, load_simchat_prompt

from chainlit import on_message, on_chat_start



from langchain.embeddings.cohere import CohereEmbeddings
from langchain.vectorstores import Pinecone
from langchain.chains import ConversationalRetrievalChain, LLMChain
from langchain.chat_models import ChatOpenAI
from langchain.memory import ConversationTokenBufferMemory
from langchain.prompts import (
    ChatPromptTemplate,
    PromptTemplate,
    SystemMessagePromptTemplate,
    HumanMessagePromptTemplate,
)
from langchain.prompts.prompt import PromptTemplate
from langchain.chains.qa_with_sources import load_qa_with_sources_chain
# from langchain.callbacks import ContextCallbackHandler

 

index_name = "chat-index-v1"

simchat = load_simchat_prompt()
query_gen_prompt = load_query_gen_prompt()
CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(query_gen_prompt)
pinecone.init(
            api_key=os.environ.get("PINECONE_API_KEY"),
            environment='gcp-starter',
    )
@cl.on_chat_start
async def on_chat_start():
    # token = os.environ["CONTEXT_TOKEN"]
    # context_callback = ContextCallbackHandler(token)
    llm = ChatOpenAI(temperature=0.0, verbose=False, openai_api_key = os.environ.get("OPENAI_API_KEY"), streaming=True, 
                     model_name='gpt-3.5-turbo',
                    #  callbacks=[context_callback]
                    )
    memory = ConversationTokenBufferMemory(llm=llm,memory_key="chat_history", return_messages=True,input_key='question',max_token_limit=1000)
    embeddings = CohereEmbeddings(model='embed-english-light-v2.0',cohere_api_key=os.environ.get("COHERE_API_KEY"))

    docsearch = Pinecone.from_existing_index(
                            index_name=index_name, embedding=embeddings
                            )
    retriever = docsearch.as_retriever(search_kwargs={"k": 4})
    
    messages = [SystemMessagePromptTemplate.from_template(simchat)]
   
    messages.append(HumanMessagePromptTemplate.from_template("{question}"))
    prompt = ChatPromptTemplate.from_messages(messages)

    question_generator = LLMChain(llm=llm, prompt=CONDENSE_QUESTION_PROMPT, verbose=False)
    doc_chain = load_qa_with_sources_chain(llm, chain_type="stuff", verbose=False, prompt=prompt)

    chain = ConversationalRetrievalChain(
            retriever=retriever,
            question_generator=question_generator,
            combine_docs_chain=doc_chain,
            verbose=False,
            memory=memory,
            rephrase_question=False,
            # callbacks=[context_callback]
        )
    cl.user_session.set("conversation_chain", chain)

    cl.user_session.set(
        "message_history",
        [{"role": "system", "content": "You are a helpful assistant."}],
    )
    # await cl.Message(content="Selamat datang! saya adalah Nada, asisten virtual anda.").send()

    
@cl.on_message
async def on_message(message: cl.Message):
    
    
    message_history = cl.user_session.get("message_history")
    message_history.append({"role": "user", "content": message.content})
        
    chain = cl.user_session.get("conversation_chain")
    
    res = await chain.arun({"question": message.content, 'stream': True})
    
    msg = cl.Message(content=res)
    message_history.append({"role": "assistant", "content": msg.content})
    await msg.send()
    await msg.update()


    print(message.content)
    print(message.id)