English | 中文

coze-discord-proxy

Proxy Discord conversations for Coze-Bot, enabling API requests to the GPT4 model with features like conversation, text-to-image, image-to-text, and knowledge base retrieval.

If you find this interesting, don't forget to give it a 🌟

📄CDP Project Documentation Site (Must-read tutorial)

🐞CDP Project - Discussion Group (Discussion)

📢CDP Project - Notification Channel (Notifications)

Features (The project's features are now stable, updates will not be frequent, feel free to raise an issue if you find bugs!)

• Perfectly compatible with NextChat, one-api, LobeChat and other conversation panels.
• Perfect support for conversation isolation.
• Conversation interface supports streaming responses.
• Supports creating discord categories/channels/threads.
• Supports conversation interface aligned with openai (v1/chat/completions) (also supports dall-e-3 text-to-image) (supports specifying discord-channel).
• Supports image-to-text/image-to-image/file-to-text interfaces aligned with openai (v1/chat/completions) (following the GPT4V interface request format [ supports url or base64 ])(supports specifying discord-channel).
• Supports dall-e-3 text-to-image interface aligned with openai (v1/images/generations).
• Supports daily 9 AM scheduled tasks to keep the bot active.
• Supports configuring multiple discord user Authorization (environment variable USER_AUTHORIZATION) for request load balancing (currently each discord user has a 24-hour limit on coze-bot calls, configure multiple users to stack request counts and balance load).
• Supports configuring multiple coze bots for response load balancing (specified through PROXY_SECRET/model), see Advanced Configuration for details.

API Documentation:

http://<ip>:<port>/swagger/index.html

Example:

How to Use

1. Open Discord's official website, log in, click settings-advanced settings-developer mode-turn on.
2. Create a discord server, right-click this server to select Copy Server ID (GUILD_ID) and record it, create a default channel in this server, right-click this channel to select Copy Channel ID (CHANNEL_ID) and record it.
3. Open Discord Developer Portal and log in.
4. Create a new application-Bot, i.e., COZE-BOT, and record its unique token and id (COZE_BOT_ID), this bot will be managed by coze.
5. Create a new application-Bot, i.e., CDP-BOT, and record its unique token (BOT_TOKEN), this bot will listen for discord messages.
6. Grant corresponding permissions (Administrator) to both bots and invite them to the created discord server (the process is not described here).
7. Open Discord's official website, enter the server, press F12 to open developer tools, send a message in any channel, find the request https://discord.com/api/v9/channels/1206*******703/messages in developer tools-Network, get Authorization (USER_AUTHORIZATION) from the header of this interface and record it.
8. Open Coze's official website, create and configure a bot (note Auto-Suggestion should be Default/on (usually no need to change)).
9. After configuration, choose to publish to discord, fill in the token of COZE-BOT, after publishing, you can see COZE-BOT online and can be used with @ in the discord server.
10. Start configuring environment variables and deploy this project using the recorded parameters.
11. Visit the API documentation address, and you can start debugging or integrating other projects.

How to Integrate with NextChat

Fill in the interface address (ip:port/domain) and API-Key (PROXY_SECRET), other fields are optional.

If you haven't set up a NextChat panel yourself, here's one already set up that you can use: NextChat

How to Integrate with one-api

Fill in BaseURL (ip:port/domain) and key (PROXY_SECRET), other fields are optional.

Deployment

Deploying with Docker-Compose (All In One)

docker-compose pull && docker-compose up -d

docker-compose.yml

version: '3.4'

services:
  coze-discord-proxy:
    image: deanxv/coze-discord-proxy:latest
    container_name: coze-discord-proxy
    restart: always
    ports:
      - "7077:7077"
    volumes:
      - ./data:/app/coze-discord-proxy/data
    environment:
      - USER_AUTHORIZATION=MTA5OTg5N************aXUBHVI  # Must modify to your discord user's authorization key (multiple keys separated by commas)
      - BOT_TOKEN=MTE5OT************UrUWNbG63w  # Must modify to the listening bot's token
      - GUILD_ID=11************96  # Must modify to the server ID where both bots are located
      - COZE_BOT_ID=11************97  # Must modify to the bot ID managed by coze
      - CHANNEL_ID=11************94  # [Optional] Default channel - (currently this parameter is only used to keep the bot active)
      - PROXY_SECRET=123456  # [Optional] API key - modify this line to the value used for request header verification (multiple keys separated by commas)
      - TZ=Asia/Shanghai

Deploying with Docker

docker run --name coze-discord-proxy -d --restart always \
-p 7077:7077 \
-v $(pwd)/data:/app/coze-discord-proxy/data \
-e USER_AUTHORIZATION="MTA5OTg5N************uIfytxUgJfmaXUBHVI" \
-e BOT_TOKEN="MTE5OTk2************rUWNbG63w" \
-e GUILD_ID="11************96" \
-e COZE_BOT_ID="11************97" \
-e PROXY_SECRET="123456" \
-e CHANNEL_ID="11************24" \
-e TZ=Asia/Shanghai \
deanxv/coze-discord-proxy

Modify USER_AUTHORIZATION, BOT_TOKEN, GUILD_ID, COZE_BOT_ID, PROXY_SECRET, CHANNEL_ID to your own values.

If the above image cannot be pulled, try using the GitHub Docker image by replacing deanxv/coze-discord-proxy with ghcr.io/deanxv/coze-discord-proxy.

Deploying to a Third-Party Platform

Deploy to Zeabur

Zeabur's servers are located abroad, automatically solving network issues, and the free tier is sufficient for personal use.

Click to deploy:

After one-click deployment, the variables USER_AUTHORIZATION, BOT_TOKEN, GUILD_ID, COZE_BOT_ID, PROXY_SECRET, CHANNEL_ID must also be replaced!

Or manually deploy:

1. First fork a copy of the code.

2. Enter Zeabur, log in with GitHub, go to the console.

3. In Service -> Add Service, choose Git (authorize first if it's your first time), select the repository you forked.

4. Deployment will automatically start, cancel it first.

5. Add environment variables

   USER_AUTHORIZATION:MTA5OTg5N************uIfytxUgJfmaXUBHVI Authorization key for discord users initiating messages (separated by commas)

   BOT_TOKEN:MTE5OTk************WNbG63w Token for the bot listening to messages

   GUILD_ID:11************96 Server ID where both bots are located

   COZE_BOT_ID:11************97 Bot ID managed by coze

   CHANNEL_ID:11************24 # [Optional] Default channel - (currently this parameter is only used to keep the bot active)

   PROXY_SECRET:123456 [Optional] API key - modify this line to the value used for request header verification (separated by commas) (similar to the openai-API-KEY)

Save.

6. Choose Redeploy.

Deploy to Render

Render provides a free tier, and linking a card can further increase the limit.

Render can directly deploy Docker images without needing to fork the repository: Render

Configuration

Environment Variables

1. USER_AUTHORIZATION=MTA5OTg5N************uIfytxUgJfmaXUBHVI Authorization key for discord users initiating messages (separated by commas)
2. BOT_TOKEN=MTE5OTk2************rUWNbG63w Token for the bot listening to messages
3. GUILD_ID=11************96 Server ID where all bots are located
4. COZE_BOT_ID=11************97 Bot ID managed by coze
5. PORT=7077 [Optional] Port, default is 7077
6. SWAGGER_ENABLE=1 [Optional] Enable Swagger API documentation [0: No; 1: Yes] (default is 1)
7. ONLY_OPENAI_API=0 [Optional] Expose only interfaces aligned with openai [0: No; 1: Yes] (default is 0)
8. CHANNEL_ID=11************24 [Optional] Default channel - (currently this parameter is only used to keep the bot active)
9. PROXY_SECRET=123456 [Optional] API key - modify this line to the value used for request header verification (separated by commas) (similar to the openai-API-KEY), recommended to use this environment variable
10. DEFAULT_CHANNEL_ENABLE=0 [Optional] Enable default channel [0: No; 1: Yes] (default is 0) If enabled, each conversation will occur in the default channel, session isolation will be ineffective, not recommended to use this environment variable
11. ALL_DIALOG_RECORD_ENABLE=1 [Optional] Enable full context [0: No; 1: Yes] (default is 1) If disabled, each conversation will only send the last content in messages where role is user, not recommended to use this environment variable
12. CHANNEL_AUTO_DEL_TIME=5 [Optional] Channel auto-delete time (seconds) This parameter is for automatically deleting the channel after each conversation (default is 5s), if set to 0 then it will not delete, not recommended to use this environment variable
13. COZE_BOT_STAY_ACTIVE_ENABLE=1 [Optional] Enable daily 9 AM task to keep coze-bot active [0: No; 1: Yes] (default is 1), not recommended to use this environment variable
14. REQUEST_OUT_TIME=60 [Optional] Non-stream response timeout for conversation interface, not recommended to use this environment variable
15. STREAM_REQUEST_OUT_TIME=60 [Optional] Stream response timeout for each stream return in conversation interface, not recommended to use this environment variable
16. REQUEST_RATE_LIMIT=60 [Optional] Request rate limit per minute per IP, default: 60 requests/min
17. USER_AGENT=Mozilla/5.0 (Macintosh; Intel Mac OS X 10_15_7) AppleWebKit/537.36 (KHTML, like Gecko) Chrome/121.0.0.0 Safari/537.36 [Optional] Discord user agent, using your own might help prevent being banned, if not set, defaults to the author's, recommended to use this environment variable
18. NOTIFY_TELEGRAM_BOT_TOKEN=6232***********Niz9c [Optional] Token for a Telegram bot used for notifications (Notification events: 1. No available user_authorization; 2. BOT_TOKEN related BOT triggers risk control)
19. NOTIFY_TELEGRAM_USER_ID=10******35 [Optional] The Telegram-Bot associated with NOTIFY_TELEGRAM_BOT_TOKEN will push to the Telegram-User associated with this variable (NOTIFY_TELEGRAM_BOT_TOKEN must not be empty if this variable is used)
20. PROXY_URL=http://127.0.0.1:10801 [Optional] Proxy (supports http only)

Advanced Configuration

Configuring Multiple Bots

1. Before deployment, create a data/config/bot_config.json file in the same directory as the docker/docker-compose deployment
2. Write the json file, bot_config.json format as follows

[
  {
    "proxySecret": "123", // API request key (PROXY_SECRET) (Note: this key must exist in the environment variable PROXY_SECRET for this Bot to be matched!)
    "cozeBotId": "12***************31", // Bot ID managed by coze
    "model": ["gpt-3.5","gpt-3.5-16k"], // Model names (array format) (if the model in the request does not match any in this json, an exception will be thrown)
    "channelId": "12***************56"  // [Optional] Discord channel ID (the bot must be in the server where this channel is located) (currently this parameter is only used to keep the bot active)
  },
  {
    "proxySecret": "456",
    "cozeBotId": "12***************64",
    "model": ["gpt-4","gpt-4-16k"],
    "channelId": "12***************78"
  },
  {
    "proxySecret": "789",
    "cozeBotId": "12***************12",
    "model": ["dall-e-3"],
    "channelId": "12***************24"
  }
]

3. Restart the service

When this json configuration is present, the bot will be matched through the [request key] carried in the request header and the [model] in the request body. If multiple matches are found, one will be randomly selected. The configuration is very flexible and can be adjusted according to your needs.

For services deployed on third-party platforms (such as zeabur) that need [configuring multiple bots], please refer to issue#30

Limitations

Current details of coze's free and paid subscriptions: https://www.coze.com/docs/guides/subscription?_lang=en

You can configure multiple discord users Authorization (refer to Environment Variables USER_AUTHORIZATION) or configure multiple bots to stack request counts and balance load.

Q&A

Q: How should I configure for high concurrency?

A: First, configure multiple bots to serve as response bots. Secondly, prepare multiple discord accounts to serve as request load and invite them into the same server, obtain the Authorization for each account, separate them with commas, and configure them in the environment variable USER_AUTHORIZATION. Each request will then pick one discord account to initiate the conversation, effectively achieving load balancing.

⭐ Star History

Related

GPT-Content-Audit: An aggregation of Openai, Alibaba Cloud, Baidu Intelligent Cloud, Qiniu Cloud, and other open platforms, providing content audit services aligned with openai request formats.

Others

Open source is not easy, if you refer to this project or base your project on it, could you please mention this project in your project documentation? Thank you!

Java: https://github.com/oddfar/coze-discord (Currently unavailable)

References

Coze Official Website: https://www.coze.com

Discord Development Address: https://discord.com/developers/applications