Forager
Turn resolved Slack threads into a queryable knowledge base.
Applied AI / agent infrastructure · Portfolio demo — hybrid sample mode plus real uploaded Slack-export backend

What it does

Forager scans resolved Slack threads, extracts canonical question-and-answer pairs, scores them for confidence, and exposes them as a queryable knowledge surface — rendered here as an MCP-style tool response.

The original demo used 10 synthetic engineering threads. It now also supports a real-data path: upload a Slack export JSON file, let the backend group it into threads and extract Q&A server-side, then query your own workspace knowledge without wiring live Slack OAuth.

What it proves

How the harvester works

  1. Question detection — identifies messages with question syntax or trailing "?".
  2. Answer resolution — scans subsequent messages for the first substantive reply with helpful reactions (💡, 👍) or confirmation signals.
  3. Confidence scoring — 0.5 base + 0.2 for resolved threads + up to 0.2 for reactions + 0.1 for explicit thanks.
  4. Query matching — token overlap scoring across question, answer, and thread context.

Honest limits

Live demo

https://dallascrilley.com/demos/forager