Skip to content

Latest commit

 

History

History
262 lines (185 loc) · 8.82 KB

README.en.md

File metadata and controls

262 lines (185 loc) · 8.82 KB
Raw

GPT AI Assistant

中文 | English

license Release

About

GPT AI Assistant is a lightweight and extensible application that is implemented using the OpenAI API and LINE Messaging API.

Through the installation process, you can start to chat with your own AI assistant using the LINE mobile app.

Demo

Installation

  • Log in to the OpenAI website.
  • Log in to the LINE website.
    • Add a provider (e.g. "My Provider").
    • Create a channel (e.g. "My AI Assistant") of type Messaging API.
    • Click the "Messaging API" tab and generate a channel access token.
  • Log in to the GitHub website.
    • Go to the gpt-ai-assistant project.
    • Click the "Star" button to support this project and the developer.
    • Click the "Fork" button to copy the source code to your own repository.
  • Log in to the Vercel website.
    • Click the "Create a New Project" button to create a new project.
    • Click the "Import" button to import the gpt-ai-assistant project.
    • Click the "Environment Variables" tab and add the following environment variables with their corresponding values:
    • Click the "Deploy" button and wait for the deployment to complete.
    • Go to the dashboard, copy the application URL, e.g. "https://gpt-ai-assistant.vercel.app/".
  • Go back to the LINE website.
    • Go to the page of "My AI Assistant", click the "Messaging API" tab, set the "Webhook URL" to application URL with "/webhook" path, e.g. "https://gpt-ai-assistant.vercel.app/webhook" and click the "Update" button.
    • Click the "Verify" button to verify the webhook call is successful.
    • Enable the "Use webhook" feature.
    • Disable the "Auto-reply messages" feature.
    • Disable the "Greeting messages" feature.
    • Scan the QR code using the LINE mobile app to add as a friend.
  • Start chatting with your own AI assistant!

Upgrade

On your own gpt-ai-assistant project page, you can click on the "Sync fork" menu and then click on either the "Update branch" or "Discard commit" button to synchronize the latest code to your repository.

When the Vercel bot detects a change in the code, it will automatically redeploy.

Commands

Send commands using the LINE mobile app to perform specific functions.

Name Alias Description
Command /command Show the application commands.
Version /version Show the application version.
Talk /talk Talk with AI Assistant.
Draw /draw Ask AI Assistant to draw a picture.
Summarize /summarize Ask AI Assistant to summarize the dialogue.
Continue /continue Ask AI Assistant to continue the conversation.
Activate /activate Activate auto-reply. The VERCEL_ACCESS_TOKEN environment variable is required.
Deactivate /deactivate Deactivate auto-reply. The VERCEL_ACCESS_TOKEN environment variable is required.
Restart /restart Deploy the application. The VERCEL_DEPLOY_HOOK_URL environment variable is required.

Environment Variables

Set environment variables to change program settings.

Name Default Value Description
APP_DEBUG false Print prompt to console. The value must be true of false.
APP_WEBHOOK_PATH /webhook Custom webhook URL path of application.
APP_LANG zh Application language. The value must be one of zh, en or ja.
SETTING_AI_NAME AI Name of AI Assistant. This is used to call AI Assistant when status is deactivated.
SETTING_AI_ACTIVATED null Status of AI Assistant. Controlled by application.
VERCEL_ACCESS_TOKEN null Vercel access token
VERCEL_DEPLOY_HOOK_URL null Vercel deploy hook URL
OPENAI_API_KEY null OpenAI API key
OPENAI_COMPLETION_MODEL text-davinci-003 Refer to model parameter for details.
OPENAI_COMPLETION_TEMPERATURE 0.9 Refer to temperature parameter for details.
OPENAI_COMPLETION_MAX_TOKENS 160 Refer to max_tokens parameter for details.
OPENAI_COMPLETION_FREQUENCY_PENALTY 0 Refer to frequency_penalty parameter for details.
OPENAI_COMPLETION_PRESENCE_PENALTY 0.6 Refer to presence_penalty parameter for details.
OPENAI_IMAGE_GENERATION_SIZE 256x256 Refer to size parameter for details.
LINE_CHANNEL_ACCESS_TOKEN null LINE channel access token
LINE_CHANNEL_SECRET null LINE channel secret

Click the "Redeploy" button to redeploy if there are any changes.

Debug

  1. Check if the environment variables of the project are filled out correctly in the Vercel.
  2. Click the "Redeploy" button to redeploy if there are any changes.
  3. If there is still a problem, please go to Issues page, describe your problem and attach a screenshot.

Development

Clone the project.

git clone [email protected]:memochou1993/gpt-ai-assistant.git

Go to the project directory.

cd gpt-ai-assistant

Install dependencies.

npm ci

Tests

Copy .env.example to .env.test.

cp .env.example .env.test

Run the tests.

npm run test

Check the results.

> [email protected] test
> jest

  console.info
    === 000001 ===
      
    Human: 嗨
    AI: OK!

Test Suites: 1 passed, 1 total
Tests:       1 passed, 1 total
Snapshots:   0 total
Time:        1 s

Using Proxy Server

Copy .env.example to .env.

cp .env.example .env

Set the environment variables as follows:

APP_DEBUG=true
APP_PORT=3000

VERCEL_GIT_REPO_SLUG=gpt-ai-assistant
VERCEL_ACCESS_TOKEN=<your_vercel_access_token>

OPENAI_API_KEY=<your_openai_api_key>

LINE_CHANNEL_ACCESS_TOKEN=<your_line_channel_access_token>
LINE_CHANNEL_SECRET=<your_line_channel_secret>

Start a local server.

npm run dev

Start a proxy server.

ngrok http 3000

Go back to the LINE website, modify the "Webhook URL" to e.g. "https://0000-0000-0000.jp.ngrok.io/webhook" and click the "Update" button.

Send a message from the LINE mobile app.

Check the results.

> [email protected] dev
> node api/index.js

=== 0x1234 ===

AI: 哈囉！
Human: 嗨？
AI: 很高興見到你！有什麼可以為你服務的嗎？

Using Docker

Copy .env.example to .env.

cp .env.example .env

Set the environment variables as follows:

APP_DEBUG=true
APP_PORT=3000

VERCEL_GIT_REPO_SLUG=gpt-ai-assistant
VERCEL_ACCESS_TOKEN=<your_vercel_access_token>

OPENAI_API_KEY=<your_openai_api_key>

LINE_CHANNEL_ACCESS_TOKEN=<your_line_channel_access_token>
LINE_CHANNEL_SECRET=<your_line_channel_secret>

Start a local server with Docker Compose.

docker-compose up -d

Changelog

Detailed changes for each release are documented in the release notes.

Credits

Related Projects

License

MIT