Skip to main content

Documentation Index

Fetch the complete documentation index at: https://docs.kb2b.app/llms.txt

Use this file to discover all available pages before exploring further.

The Knowledge section (/dashboard/knowledge) is the main panel for understanding the POT’s epistemic health: how reliable is what it knows? what needs human attention? what’s moved recently? (In earlier versions this was called Memory — the rename to Knowledge happened in v0.14.0. If you see “Memory” in an old screenshot, it’s the same thing.)

The 5 Trust Dashboard signals

The main page has 5 cards all answering, in one place, “can I trust what’s in here?”:

1. Contradictions (banner)

A persistent banner at the top of the section showing how many pending contradictions there are and their severity (critical / moderate / mild). Click a contradiction → opens the resolution flow directly (see POT Constitution). If there are no contradictions, the banner hides itself. That’s a good signal.

2. Last-7-days engagement

A compact strip with:
  • Number of actions on facts in the last 7 days (approvals, disapprovals, comments)
  • Active users in the workspace
  • Delta vs previous week (green if more, red if less)
  • An inline sparkline showing the day-by-day trend
“See detail” jumps to the Activity sub-route (below).

3. Validation coverage

% of facts that have received at least 1 validation (approval or disapproval).
  • Green / “healthy” if ≥ 70%
  • Orange / “at risk” if below 70%
It’s the metric for “what fraction of the knowledge has had human eyes on it?”. If it’s low, your team is trusting too much of what kb2b auto-extracts without verifying.

4. Freshness

% of facts touched (validated, edited, or referenced in chat) in the last 30 days. Together with validation coverage, it tells you not only “has it been seen?” but also “is it still current?”. A fact validated 2 years ago might be obsolete by now.

5. Needs review

A queue of the top 10 unvalidated facts, sorted by ascending POT Score (most uncertain first). The actionable list of “sit down for 15 minutes and review these”. Each fact in the queue can be approved/disapproved inline, or you can open the inspector for more detail.

Recent evidence

Below the 5 cards, a Recent evidence section showing the last 4 facts added to the POT. Useful to see “what got ingested this week” at a glance.

Sub-routes

The Knowledge section has 3 sub-routes with their own sub-nav:
RouteWhat’s in it
/dashboard/knowledgeThe Trust Dashboard described above (5 cards + recent evidence)
/dashboard/knowledge/activityFull engagement widget (7d / 30d / all-time filters) + recent facts + activity feed
/dashboard/knowledge/constitutionConstitutional axioms editor + active campaigns + danger zone (delete POT). See POT Constitution for detail.
The contradictions banner appears on all sub-routes — it’s global to the section. You can share specific URLs with your team:
  • /dashboard/knowledge?fact={id} → opens the inspector for a specific fact
  • /dashboard/knowledge?contradiction={id} → opens a specific contradiction for resolution
  • /dashboard/knowledge/constitution?contradiction={id} → takes you to the constitution editor with the contradiction already selected
Old URLs like /dashboard/pot or /dashboard/activity still work — they auto-redirect (308) to the new Knowledge routes.

How to use this day-to-day

Monday morning (10 minutes):
  1. Check the contradictions banner. If there are critical ones, resolve those first.
  2. Review the engagement delta vs the previous week. Did team activity on the POT go up or down?
  3. Coverage and freshness: are they above 70% and 50% respectively? If not, focus the team this week on validating.
  4. Needs review: approve/disapprove 5-10 facts before starting anything else.
In team meetings: use the Activity tab to show “what we moved on account knowledge this week”. Before a quarter: review coverage + freshness as POT health metrics. Below certain thresholds, decide whether to pause new ingestions until you clean up.

When NOT to look at this

  • If your POT has fewer than 20 facts, the metrics aren’t meaningful. Get it to at least 50 facts (by uploading documents and recording meetings) before reading trust signals seriously.
  • If you’re a solo user (no team), the Engagement and Needs review cards offer little — it’s just you. Focus on chat and meetings.
Page in progress Phase 1 — screenshots of the 5 cards and sub-routes coming in Phase 2.