A Slack bot powered by GPT
Tireless Assistant
Invite Tireless Assistant into a channel and ask it to summarize all the previous messages you've missed via its slash command.
I've written a Slack bot which I'm calling Tireless Assistant. It's purpose is to quickly catch you up on the latest in a channel without you having to read through a litany of old messages.
Usage
Use the slash command /sum to have Tireless Assistant read and summarise the last 6 hours.
Add custom time spans to change how far back it reads.
Add publicly to the end of the command to post the summary in the channel for all to see.
Examples
/sum 1h30m will summarise the past 1 hour and 30 minutes of messages.
/sum 1 day 2 hours will summarise the past 26 hours.
/sum 45 min publicly will summarise the past 45 minutes of messages and post that summary in the channel.
FAQ
How does it work?
The mechanism is pretty simple:
- Parse the incoming slash command for a timespan (e.g. "2 days", "3h 30 min")
- Fetch all the messages between now and that amount of time in the past
- Format it in a way that can be easily interpreted
- Send it to GPT and ask it to summarise it
- Privately post that summary to the requesting user
Can Tireless Assistant read my private channels or DM's?
No. Tireless Assistant can only read messages in channels or conversations it's been added to. If someone uses /sum in a private conversation, it will do nothing as the bot cannot join a private channel or DM group by itself. If you or a member explicitly invites Tireless Assistant in, the bot will work as usual.
Does Tireless Assistant store any information from my Slack?
No. The bot is stateless. Any information it reads only exists for the life of that http request. The only stored value is an access token granted by Slack when you install the app. The bot uses this token to communicate with Slack so it can read what it needs to summarise and subsequently send the summary.
The full source code is available on the bot's repo on my Github.
After you ask GPT to summarise something, will it store and use that for training?
As stipulated in OpenAI's Terms of Use, OpenAI will not use any content from their API requests to train their language models. See: 3. Content - (c) Use of Content to Improve Services. This is how Tireless Assistant communicates with GPT.
Will Tireless Assistant always be free?
Probably. However GPT is not a free service. If the bot becomes very popular and the OpenAI API bills get unmanageable, I may seek to monetise it for enterprise orgs.
Privacy Policy
You can read the full privacy policy here
