SlackFlow AI

Inspiration## Inspiration

Every team's most valuable knowledge — decisions, action items, code snippets, deadlines — gets buried in endless Slack threads within days. We wanted a way to turn everyday conversations into a structured, searchable project workspace without anyone having to manually take notes.

What it does

SlackFlow AI is an AI-powered Slack assistant that:

• Uses /capture to extract action items, decisions, documentation links, code snippets, and deadlines from any Slack thread (via permalink) or pasted conversation text, using an LLM.
• Uses /flow-find to search everything ever captured, with results shown in rich Block Kit modals.
• Provides a live dashboard (/flow-dashboard and the App Home tab) showing open action items, decisions logged, and upcoming deadlines.
• Runs automated daily standups — posting a reminder, collecting Yesterday/Today/Blockers via a modal form, and posting an AI-generated digest that explicitly calls out blockers.
• Tracks action item status with interactive buttons (mark as done).

How we built it

• Backend: Python + Flask, using the Slack Bolt SDK for all Slack event/command/interactivity handling.
• Database: SQLite, with a schema covering workspaces, users, captures, action items, decisions, doc links, code snippets, deadlines, and standups.
• AI: OpenAI-compatible chat completions API for structured knowledge extraction and standup summarization, with a regex-based fallback extractor so the bot degrades gracefully instead of failing if the AI call errors out.
• UI: Slack Block Kit for all modals, dashboards, and interactive messages.
• Scheduling: APScheduler for daily standup reminders and digests.

Challenges we ran into

• Getting Slack permalink parsing right (decoding the timestamp format) to fetch full thread context via conversations.replies.
• Designing prompts that reliably return strict, parseable JSON from the LLM for extraction.
• Handling Slack's view/modal update flow correctly (ack vs. response_action) for a smooth search & capture experience.

Accomplishments that we're proud of

A fully working end-to-end pipeline — from a raw Slack thread to structured, searchable knowledge — with proper error handling, logging, and a fallback path so the bot never silently fails.

What we learned

Deep familiarity with the Slack Bolt SDK's event/command/view lifecycle, Block Kit UI design constraints, and prompt engineering for reliable structured extraction.

What's next for SlackFlow AI

• Postgres support for multi-workspace scale
• Slash command for editing/reassigning action items directly
• Smarter deadline reminders integrated with Slack's reminder API

What it does

How we built it

Challenges we ran into

Accomplishments that we're proud of

What we learned

What's next for SlackFlow AI

Built With

• apscheduler
• flask
• ngrok
• openai
• python
• slack-block-kit
• slack-bolt
• sqlite