Summarize conversations that you don't have time to read.
/tldr <message_link>: Summarizes the conversation following the message linked.
/tldr https://discord.com/channels/1234567890/0987654321/1122334455
The bot will also produce periodic summaries of server messages, outputting results to a specified channel. The ID of the channel must be provided in the SUMMARY_OUTPUT_CHANNEL_ID environment variable.
Summaries can be deactivated and re-activated at will using the /deactivate_summary and /activate_summary commands.
- Python (known to work with 3.11)
- Discord Bot Token (How to create a Discord Bot)
- OpenAI API Key (OpenAI API)
-
Clone the Repository:
git clone https://github.com/leofdgit/discord-tldr-bot.git cd discord-tldr-bot -
Install Dependencies:
pip install -r requirements.txt- Setup Environment Variables:
Edit the .env file in the project directory to populate it with your Discord Bot Token and OpenAI API Key. See config for details on how these environment variables influence the bot's behaviour.
ENV_FILE=/path/to/env/file python start.pyDocumentation on the various configuration options available (as environment variables).
Comma-separated Discord user IDs. If supplied, only users whose IDs are included here will be permitted to use the bot's commands.
AUTHORIZED_USER_IDS=12345,67890
The Discord bot token, which grants the application access to Discord. See how to create a Discord Bot.
DISCORD_CLIENT_KEY=abcde
The ID of the Discord guild (or "server") that the bot should interact with.
GUILD_ID=xyz
Note: it is possible for a single Discord bot to server multiple servers. Restricting this bot to work for a single server is a self-imposed restriction in order to not have to worry about accidentally leaking messages across servers.
When producing periodic summaries, any channels with fewer than SUMMARY_MSG_LOWER_LIMIT new messages in the summary
period will be ignored. Defaults to 0, meaning all channels will be included in periodic summaries.
SUMMARY_MSG_LOWER_LIMIT=0
Periodic summaries will be sent to this channel.
SUMMARY_OUTPUT_CHANNEL_ID=12345
Periodic summaries will be produced every SUMMARY_INTERVAL seconds. Defaults to one day in seconds.
SUMMARY_INTERVAL=86400
Accepts values "true" and "false". If set to "true" then periodic summaries will begin immediately once
the bot launches. If set to "false" then periodic summaries need to be kick started using the /activate_summary
command. Defaults to "false".
SUMMARY_AUTOSTART=false
No more than MAX_OUTPUT_TOKENS tokens will be used to generate the summary. Defaults to 200 as this seems to be a sufficient upper bound. See What are tokens?.
MAX_OUTPUT_TOKENS=200
The company that provides the AI model that will be used. Can be thought of as a namespace for AI_MODEL.
At present, only a single value is supported: open_ai. Defaults to open_ai.
AI_PROVIDER=open_ai
The model that should be used to generate periodic summaries and tl;drs. Typically, the more expensive the model, the better the summaries. See OpenAI Models.
AI_MODEL=gpt-3.5-turbo
The secret value that gives your bot instance the ability to send queries to the AI provider. See (OpenAI API).
AI_API_KEY=wasd
- AI Providers: currently only supports OpenAI models.
- API Rate Limits: be mindful of OpenAI's API usage limits.
- Cost: GPT4 is quite expensive; at the time of writing, summarizing 200 short-ish messages incurred a cost of $0.10. GPT3.5 is much cheaper.
- Behaviour: the summaries given are not deterministic, meaning that summarizing the same content twice may give two different results. The summarizies may be innacurate.
- Privacy Considerations: ensure users in your Discord server are aware that their messages may be processed by this bot, and hence the AI provider, for summarization.
... are welcome! Please raise an issue to get the ball rolling!