This a Collaborative Automation Liaison (CAL). Cal is mostly an exploration on things we can do in Slack with an agentic interface. Today it can
- Search Slack and summarize the responses
- Summarize a thread
- Recommend workflows to accomplish tasks
Cal is activated either by mentioning the app (
@cal) or starting a DM/Agent session.
If you have ideas for how we can make Cal more useful, send a message to @Naseer Rashid in Slack.
This app uses [https://api.slack.com/apis/socket-mode](socket mode). Here are the deploy methods that have been tested:
- Local dev: Socket mode, https://docs.python.org/3/library/venv.html, and a B+ workspace
- Cloud deploy: Heroku, Socket mode, Slack.0 grid environment.
SLACK_USER_TOKEN: `User OAuth Token`
BOT_USER_TOKEN: `Bot User OAUTH Token`
SLACK_APP_TOKEN=`App-Level Token`
OPENAI_API_KEY=`Your OpenAI Key`
SLACK_TEAM_ID=`Workspace ID` - Starts with a T. In grid, just pick a workspace your user is already in. Doesn't really matter which one.
SLACK_ENTERPRISE_ID=`Grid Enterprise ID` - Leave blank for single-workspace deploys
Call uses a combination of OpenAI models. Feel free to consolidate to your favorite one.
Most limitations are related to rate limiting on the LLM side. DM experience in a single-workspace environment doesn't always work how we'd expect an agent session to behave.
Installing this app for multiple users in an environment hasn't been tested so do it at your own risk. There's no token managemet mechanism built into this app.