Skip to content

Latest commit

 

History

History
216 lines (148 loc) · 9.61 KB

README.md

File metadata and controls

216 lines (148 loc) · 9.61 KB
Raw

ChatGPT Line Bot

🤖Introduction

Integrate ChatGPT Bot into Line, simply enter text in the input box to start interacting with ChatGPT.

✨Features

  • Completely free ChatGPT Bot

  • Weekly horoscope information (real-time)

  • Scheduled push notifications for YT music channel

  • Online image search (real-time)

Note

If there're any new feature requirements, welcome to submit a PR or ISSUE

🔨Tools

  • Python FastAPI: Build ChatGPT response API
  • gpt4free: Free to use OpenAI API
  • Line messaging API channel: Connect to ChatGPT API
  • Github: Store the code
  • replit: Free deployment of your own FastAPI
  • CronJob: Free scheduled requests to prevent API interruptions
  • render, ngrok: Other free deployment alternatives

🎈Installation Steps

Token Acquisition

  1. Obtain Line Token:
    1. Log in to Line Developer
    2. Create a bot:
      1. Create Provider -> Click Create
      2. Create Channel -> Choose Create a Messaging API channel
      3. Enter the required basic information
      4. After entering the information, under Basic Settings, there is a Channel Secret -> Click Issue, the generated value is LINE_CHANNEL_SECRET (will be used later)
      5. Under Messaging API, there is a Channel access token -> Click Issue, the generated value is LINE_CHANNEL_ACCESS_TOKEN (will be used later)

Project Setup

  1. Fork the Github project:
    1. Register/Login to GitHub
    2. Go to ChatGPT-Line-Bot
    3. Click Star to support the developer
    4. Click Fork to copy all the code to your own repository
  2. Deployment (Free Space):
    1. Go to replit
    2. Click Sign Up and log in with your Github account and authorize it -> Click Skip to skip the initial setup
    3. On the main page in the middle, click Create -> In the popup box, click the top right corner Import from Github
    4. If you haven't added your Github repository, click the link Connect GitHub to import your private repos. -> Check Only select repositories -> Choose ChatGPT-Line-Bot
    5. Go back to step 4, at this time the Github URL can choose the ChatGPT-Line-Bot project -> Click Import from Github.

Project Execution

  1. Environment Variable Setup

    1. After completing the previous step Import in Replit, click Tools at the bottom left of the project management page.

    2. After clicking Got it on the right side, you can add environment variables, you need to add:

      1. Line Channel Secret:
        • key: LINE_CHANNEL_SECRET
        • value: [obtained from step 1]
      2. Line Channel Access Token:
        • key: LINE_CHANNEL_ACCESS_TOKEN
        • value: [obtained from step 1]

  2. Start Execution

    1. Click Run at the top
    2. After successful execution, the right side will display Hello World, and copy the URL in the upper right corner of the screen
    3. Go back to Line Developer, in Messaging API, paste the URL from above, and add /callback to it, for example: https://ChatGPT-Line-Bot.jimmylin.repl.co/callback
    4. Turn on Use webhook
    5. Turn off Auto-reply messages below

    • Note: If there are no requests within an hour, the program will be interrupted, so the next step is needed

  3. CronJob Scheduled Requests

    1. Register/Login to cron-job.org
    2. After logging in, click CREATE CRONJOB in the upper right corner of the panel
    3. Enter ChatGPT-Line-Bot in Title, enter the URL from the previous step, for example: https://ChatGPT-Line-Bot.jimmylin.repl.co/
    4. Below, set it to run every 5 minutes
    5. Click CREATE

Connect Service and Line Bot

Go back to Line Developer homepage and click Join friend guide to scan the LINE Bot QR code:

Homepage -> Click your bot -> Click Add friends tool -> Create friend action barcode (https://manager.line.biz/account//gainfriends)

Congratulations on creating your first LINE Bot! Try talking to it and see how it responds!

⛄ Group and Non-Group

  • When chatting with the bot individually, any message will trigger a response.
  • In a group, to interact with the bot, you need to use the @chat command at the beginning of your message. For example: @chat hi~

🎃 Special Features

Horoscope

When your message includes a request for horoscope information, it triggers a web crawler to fetch the weekly horoscope:

  • Personal chat: Give me the Scorpio horoscope, I want to know the Scorpio horoscope, ...
  • Group chat: @chat Give me the Scorpio horoscope, @chat I want to know the Scorpio horoscope, ...

Online Image Search

When your message includes a request to search for images, it triggers a web crawler to fetch the desired image:

  • Personal chat: Search online for a picture of Lin Xiang smoking, Find me a picture of Lin Xiang smoking online, ...
  • Group chat: @chat Search online for a picture of Lin Xiang smoking, @chat Find me a picture of Lin Xiang smoking online, ...

📢Broadcast Message - Daily Youtube Song Recommendation

  • With the broadcast API, we can send messages to every user at once using Line Bot
  • Here we want the Line Bot to randomly recommend 3 good Youtube songs every morning:

    • Create the data ./data/favorite_videos.json, you can refer to the author's data

      (The data is created by fetching the author's favorite videos using the Youtube Data v3 API, which is not specifically introduced here)

    • Use ./chatgpt_linebot/modules/youtube_recommend.py to randomly select 3 songs and have GPT organize them

    • Add /recommend route in ./chatgpt_linebot/urls.py:

      videos = recommend_videos() # Get 3 songs

if videos:
    line_bot_api.broadcast(TextSendMessage(text=videos)) # Use broadcast to send messages to users
    
    # Since broadcast cannot send push messages in groups, you can push messages to specified group ids through known group ids
    # You can ignore the code below if you don't need to send messages to specific groups
    known_group_ids = [
        'C6d-xxxxxxxxxxxxxxxxxxxxxxxxxxxxx',
        'Ccc-xxxxxxxxxxxxxxxxxxxxxxxxxxxxx',
        'Cbb-xxxxxxxxxxxxxxxxxxxxxxxxxxxxx',
    ]
    for group_id in known_group_ids:
        line_bot_api.push_message(group_id, TextSendMessage(text=videos))

      To get the group_id of a group, you can test it in the replit console using print:

      elif event.source.type == 'group' and user_message.startswith('@chat'):
    group_id = event.source.group_id
    print(group_id) # Output group_id
    memory.append(group_id, 'user', refine_message.replace('@chat', ''))
    response = chat_completion(group_id, memory)

    • Now, when we access the /recommend route, it will trigger the message broadcast, and all users and specified groups will receive the message

    • Next, we use cron-job.org again for scheduling. Set it to call this API every morning at 8:00 to achieve daily push notifications!

⚔Advanced - Personalized Bot

In addition, we can use the prompt method to allow the Line Bot to answer in a personalized way. In ./chatgpt_linebot/prompts/template.py, we can define a template, for example:

Question: What did you have for breakfast today?

Bot Answer: Darling, did you wake up this morning? I've been waiting for you in bed, and I'm so hungry just thinking about your body. What should we have for breakfast today? How about a spicy omelette, as hot as your charming figure? 😏🍳

Free Deployment Alternatives

Since replit no longer supports a free plan, the author provides the following two alternative solutions:

  • render: The deployment method is similar to replit, and will not be extensively discussed here. Although this solution is Serverless, it tends to be unstable.
  • ngrok: Use your local computer as a server to deploy the API.

    • Download ngrok for your operating system.

    • Add the path to ngrok.exe to your environment variables.

    • In the Terminal, start FastAPI with: pip install poetry

      poetry shell

      poetry install

      $env:LINE_CHANNEL_SECRET="..."; $env:LINE_CHANNEL_ACCESS_TOKEN="..."; $env:SERPAPI_API_KEY="..."; poetry run python main.py

    • In the Terminal, run: ngrok config add-authtoken <token>, where the token is from your personal account on the ngrok website.

    • In the Terminal, execute the following commands to start ngrok:

      tskill /A ngrok
ngrok http 8080

    • Use the Forwarding URL as your Web URL.

      ngrok forwarding

Finally, remember to replace the obtained URL in the Webhook URL section under Messaging API in Line Developer.

References

  1. Line_Bot_Tutorial

  2. ChatGPT-Line-Bot

Back to top