langchain_mem.py

import os
import sys

import aiohttp

from fastapi import Request, FastAPI, HTTPException

from langchain.chat_models import ChatOpenAI
from langchain.chains import ConversationChain
from langchain.memory import ConversationBufferWindowMemory


from linebot import (
    AsyncLineBotApi, WebhookParser
)
from linebot.aiohttp_async_http_client import AiohttpAsyncHttpClient
from linebot.exceptions import (
    InvalidSignatureError
)
from linebot.models import (
    MessageEvent, TextMessage, TextSendMessage,
)

import os
import openai

from dotenv import load_dotenv, find_dotenv
_ = load_dotenv(find_dotenv())  # read local .env file

# get channel_secret and channel_access_token from your environment variable
channel_secret = os.getenv('ChannelSecret', None)
channel_access_token = os.getenv('ChannelAccessToken', None)
if channel_secret is None:
    print('Specify LINE_CHANNEL_SECRET as environment variable.')
    sys.exit(1)
if channel_access_token is None:
    print('Specify LINE_CHANNEL_ACCESS_TOKEN as environment variable.')
    sys.exit(1)

app = FastAPI()
session = aiohttp.ClientSession()
async_http_client = AiohttpAsyncHttpClient(session)
line_bot_api = AsyncLineBotApi(channel_access_token, async_http_client)
parser = WebhookParser(channel_secret)

# Langchain 串接 OpenAI ，這裡 model 可以先選 gpt-3.5-turbo
llm = ChatOpenAI(temperature=0.9, model='gpt-3.5-turbo')

# 透過 ConversationBufferWindowMemory 快速打造一個具有「記憶力」的聊天機器人，可以記住至少五回。
# 通常來說 5 回還蠻夠的
memory = ConversationBufferWindowMemory(k=5)
conversation = ConversationChain(
    llm=llm,
    memory=memory,
    verbose=False
)


@app.post("/callback")
async def handle_callback(request: Request):
    signature = request.headers['X-Line-Signature']

    # get request body as text
    body = await request.body()
    body = body.decode()

    try:
        events = parser.parse(body, signature)
    except InvalidSignatureError:
        raise HTTPException(status_code=400, detail="Invalid signature")

    for event in events:
        if not isinstance(event, MessageEvent):
            continue
        if not isinstance(event.message, TextMessage):
            continue

        # 將使用者傳來的訊息 event.message.text 當成輸入，等 LangChain 傳回結果。
        ret = conversation.predict(input=event.message.text)

        await line_bot_api.reply_message(
            event.reply_token,
            TextSendMessage(text=ret)
        )

    return 'OK'