Abstract

LineGPT - ChatGPT Integration for LINE

This bot integrates ChatGPT with LINE messaging platform, allowing users to interact with ChatGPT through LINE messages.

Features

• Chat with ChatGPT through LINE messages
• Support for conversation history
• Easy setup and deployment
• Configurable ChatGPT parameters
• Daily financial news updates
• Command system for different interactions

🔄 LineGPT Workflow

The following diagram illustrates the architecture and data flow of the LineGPT application:

flowchart LR
    User(LINE User) <-->|Messages| LINE[LINE Platform]
    LINE <-->|Webhook Events| Server[LineGPT Server]
    Server -->|User Query| Process[Message Processing]
    Process -->|Command Handling| Commands{Commands}
    Process -->|Regular Message| Chat[Chat Processing]
    
    Commands -->|/chat| Chat
    Commands -->|/reset| Reset[Reset Conversation]
    Commands -->|/news| News[Financial News]
    Commands -->|/help| Help[Command List]
    
    Chat -->|API Request| OpenAI[OpenAI GPT API]
    OpenAI -->|AI Response| Chat
    
    News -->|API Request| FinAPI[Financial News API]
    FinAPI -->|News Data| News
    
    Chat -->|Format Response| Response[Format Response]
    News -->|Format News| Response
    Reset -->|Confirmation| Response
    Help -->|Command Info| Response
    
    Response -->|LINE Message| Server
    
    subgraph "Debug Endpoints"
    Debug[Debug Routes]
    Server -.->|Testing| Debug
    Debug -.->|Raw Message| LINE
    Debug -.->|Test Encoding| OpenAI
    end
    
    style User fill:#f9d71c,stroke:#333,stroke-width:2px
    style LINE fill:#00c300,stroke:#333,stroke-width:2px,color:#fff
    style Server fill:#5762d5,stroke:#333,stroke-width:2px,color:#fff
    style OpenAI fill:#74aa9c,stroke:#333,stroke-width:2px,color:#fff
    style FinAPI fill:#ff6b6b,stroke:#333,stroke-width:2px
    style Debug fill:#ff9e00,stroke:#333,stroke-width:2px

This diagram shows how user messages flow through the LINE platform to the LineGPT server, where they are processed based on command type, then routed to appropriate services (OpenAI API or Financial News API), and finally formatted and sent back to the user through LINE.

Testing Results

Example Test Commands

Here are some examples of how to test your LineGPT setup:

# Test sending a simple message to a user
curl "http://localhost:5000/test_message?user_id=USER_ID_HERE&message=Hello from LineGPT!"

# Test sending Chinese and other non-ASCII messages
curl "http://localhost:5000/test_chinese?user_id=USER_ID_HERE"

# Test sending a message to all registered users
curl "http://localhost:5000/test_broadcast?message=Hello everyone from LineGPT!"

# Test encoding with specific text
curl "http://localhost:5000/test_encoding?user_id=USER_ID_HERE&text=测试中文消息"

Replace USER_ID_HERE with your actual LINE user ID, which you can obtain from the /get_user_id endpoint.

What You Can Ask GPT

With LineGPT, you can ask GPT about virtually anything, including:

• General knowledge questions
• Help with writing or composing messages
• Explanation of complex concepts
• Financial advice and market insights
• Programming help and code examples
• Language translations
• Creative writing prompts
• Math problems and calculations
• Historical information
• Current events (up to its knowledge cutoff)
• Recommendations for books, movies, etc.

Simply start your message with a question or statement, or use the /chat command followed by your query.

Example Interactions

Here are some examples of how you can interact with GPT through LINE:

General Questions

You: What is the capital of France?
GPT: The capital of France is Paris.

Language Translation

You: /chat Translate "Hello, how are you?" to Chinese
GPT: 你好，你好吗？(Nǐ hǎo, nǐ hǎo ma?)

Programming Help

You: /chat Write a Python function to calculate Fibonacci numbers
GPT: Here's a Python function to calculate Fibonacci numbers:

def fibonacci(n):
    if n <= 0:
        return 0
    elif n == 1:
        return 1
    else:
        return fibonacci(n-1) + fibonacci(n-2)

# More efficient version using dynamic programming
def fibonacci_efficient(n):
    fib = [0, 1]
    for i in range(2, n+1):
        fib.append(fib[i-1] + fib[i-2])
    return fib[n]

Financial Analysis

You: /chat What factors are affecting the tech stock market right now?
GPT: Several factors currently affecting the tech stock market include:

1. Interest rate changes by central banks
2. Inflation concerns
3. Earnings reports from major tech companies
4. AI development and adoption rates
5. Regulatory changes and antitrust investigations
6. Supply chain challenges
7. Consumer spending patterns
...

GPT's Capabilities

GPT-4 and similar models have impressive capabilities that make them useful for many different tasks:

1. Natural Language Understanding

GPT can understand nuanced questions and provide relevant, contextual responses.

2. Multi-lingual Support

It can communicate in dozens of languages, including Chinese, Japanese, Korean, Spanish, French, and many more.

3. Knowledge Base

The model contains knowledge about a wide range of topics up to its training cutoff date.

4. Code Generation and Analysis

GPT can write code in many programming languages, debug existing code, and explain how code works.

5. Creative Writing

It can help with writing stories, poems, essays, emails, and other forms of creative or professional writing.

6. Logical Reasoning

GPT can follow complex instructions and solve problems step by step.

7. Limitations to Be Aware Of

• GPT doesn't have access to the internet (unless specifically connected)
• Its knowledge has a cutoff date and may not include very recent events
• It may occasionally provide incorrect information
• It cannot execute code or perform actions outside of generating text responses

Prerequisites

• Python 3.8 or higher
• LINE Messaging API channel
• OpenAI API key
• Financial Modeling Prep API key (for financial news)
• ngrok account (free) for webhook setup

Setup

1. Clone this repository

git clone https://github.com/yourusername/LineGPT.git
cd LineGPT

2. Create a virtual environment and install dependencies

python3 -m venv venv
source venv/bin/activate
pip install -r requirements.txt

3. Create a .env file in the root directory with your credentials:

LINE_CHANNEL_ACCESS_TOKEN=your_line_channel_access_token
LINE_CHANNEL_SECRET=your_line_channel_secret
OPENAI_API_KEY=your_openai_api_key
FINANCIAL_NEWS_API_KEY=your_financial_news_api_key
USER_IDS=user_id1,user_id2  # Comma-separated list of LINE user IDs to send news to

4. Run the bot

# Option 1: Using the run script
./run.sh

# Option 2: Manually activating the virtual environment
source venv/bin/activate
python3 app.py

5. Set up the webhook URL

# In a new terminal window, run the webhook setup script
./setup_webhook.sh

This script will:

• Install ngrok if it's not already installed
• Prompt you for your ngrok authtoken if not configured
• Create a tunnel to your local server
• Display the webhook URL to use in LINE Developers Console

Getting LINE Credentials

1. Go to LINE Developers Console
2. Create a new provider (if you haven't already)
3. Create a new channel of type "Messaging API"
4. Get your Channel Secret and Channel Access Token from the channel settings

Setting Up ngrok for Webhook

1. Sign up for a free ngrok account at https://dashboard.ngrok.com/signup
2. After signing up, go to https://dashboard.ngrok.com/get-started/your-authtoken
3. Copy your authtoken
4. When you run ./setup_webhook.sh, you'll be prompted to enter this authtoken if ngrok is not already configured

Setting Up LINE Webhook

1. Run your LineGPT application using ./run.sh
2. In a separate terminal, run ./setup_webhook.sh to start ngrok
3. Enter your ngrok authtoken when prompted (if not already configured)
4. Copy the webhook URL provided by the script
5. Go to LINE Developers Console
6. Select your provider and channel
7. Go to the "Messaging API" tab
8. Scroll down to "Webhook settings"
9. Enter the webhook URL (ending with /callback) in the "Webhook URL" field
10. Click "Update" and then "Verify"
11. Make sure "Use webhook" is enabled (set to "Enabled")
12. Optionally, turn off "Auto-reply messages" to let your bot handle all messages

Getting OpenAI API Key

1. Go to OpenAI API Keys
2. Create a new API key
3. Copy the key and add it to your .env file

Getting Financial News API Key

1. Go to Financial Modeling Prep
2. Sign up for an account
3. Get your API key and add it to your .env file

Usage

1. Add the bot as a friend on LINE using the QR code from your LINE channel
2. Send a message to the bot to chat with ChatGPT
3. Use commands for specific functions:
   • /chat [message] - Explicitly ask ChatGPT a question
   • /reset - Clear conversation history
   • /news - Get the latest financial news on demand
   • /help - Show available commands

Configuration

You can modify the following parameters in .env:

• Model selection (e.g., gpt-4, gpt-3.5-turbo)
• Maximum tokens per response
• Temperature setting
• System prompt
• News delivery time

Deployment

For production deployment:

1. Use a proper WSGI server like Gunicorn
2. Set up HTTPS for your webhook URL
3. Consider using a process manager like Supervisor

To expose your local development server for testing:

• Use the included setup_webhook.sh script
• Or manually run: ngrok http 8080
• Set the webhook URL in LINE Developers Console to your ngrok URL + "/callback"

Troubleshooting

Webhook Issues

If you're experiencing issues with the webhook:

1. Make sure your ngrok tunnel is running and the webhook URL is correctly set in the LINE Developers Console.
2. Check that the Channel Secret and Channel Access Token are correctly set in your .env file.
3. Verify that your server is running and accessible from the internet.
4. Check the logs for any error messages.

Debugging Endpoints

The application includes several debugging endpoints to help diagnose issues:

• /debug - Provides general debugging information about the application, including encoding settings and API key status.
• /test_message?user_id=USER_ID&message=MESSAGE - Sends a test message to a specific user.
• /test_chinese?user_id=USER_ID - Tests sending messages in multiple languages (English, Chinese, Japanese, Korean).
• /raw_message?user_id=USER_ID&message=MESSAGE - Sends a raw message directly to the LINE API, bypassing the application's message handling.
• /test_broadcast?message=MESSAGE - Sends a message to all registered users.
• /test_encoding?user_id=USER_ID&text=TEXT - Tests text encoding and sends a message.

These endpoints can be accessed when the application is running, for example:

http://localhost:5000/debug
http://localhost:5000/test_message?user_id=YOUR_USER_ID&message=Test message

Bot Response Issues

If the bot is not responding to messages:

1. Check that your OpenAI API key is valid and has sufficient credits.
2. Verify that the LINE Channel Access Token has the correct permissions.
3. Check the application logs for any error messages.
4. Try using the /test_message endpoint to send a test message directly.
5. For issues with Chinese or other non-ASCII text, try the /test_chinese endpoint.

LibreSSL Warnings

If you see warnings related to LibreSSL:

urllib3.exceptions.InsecureRequestWarning: Unverified HTTPS request is being made to host

This is a warning and not an error. The application should still function correctly. To resolve this warning:

1. Upgrade to Python 3.10 or later.
2. Or, install the pyopenssl package: pip install pyopenssl

License

MIT License

Methodology

Results