About Services Tools Blog Contact us

Keeping the Knowledge Base in Step with Its Sources

A scheduled job that will pull changes from each connected source, refresh who may read what, and hide items removed at the source. The live connections are still to be built.

  • Scheduled Workers & Sync
  • Professional services
  • Scheduled worker
  • Product
  • Local/private
  • 2026

The problem

A private knowledge base goes stale, or keeps showing documents the source system has already removed or locked down. People then answer from a passage they should no longer see. Rebuilding everything by hand is too slow to do every day.

What it does

The sync visits each source and picks up from where it last stopped, saving new or changed passages. Access lists are updated to match the source. Items deleted at the source are marked as removed, so later answers skip them. This build is a framework only. The schedule, the safety check and the data layout are in place, and the live connections are still to be built.

Outcomes

  • Once a source is connected, the knowledge base follows it instead of relying on a one-off import.
  • Removed or locked documents can be hidden so answers skip them.
  • The design includes a health status and a saved position for each source. The job does not write them yet.

Simplified mockup

Interactive preview with sample data. Nothing is sent to a server.

Knowledge connector sync

Scaffold
Saved cursors 10 Oct 2026, before 07:15
ConnectorCursorHealthPreset
Files 10 Oct 2026, 06:00 Synced Quiet this hour
Mail 9 Oct 2026, 18:12 Due
Wiki 8 Oct 2026, 21:40 Due Changes waiting
This run
Schedule is in place. Live calls are still to be filled in.

Passages

Index
TitleSourceReaders
Office address and opening hoursFilesPractice
How to book the meeting roomWikiPractice
Supplier payment termsWikiAccounts

Connector health

Last run · 9 Oct 2026, 22:10
ConnectorCursorHealthNote
Files 10 Oct 2026, 06:00 Healthy 0 new since the hour
Mail 9 Oct 2026, 18:12 Healthy Caught up last night
Wiki 8 Oct 2026, 21:40 Due Cursor older than a day
Passages added this week

The schedule, the command-line guard and the schema are in place. Live connector calls are still to be filled in. This desk shows what that run will write.

How it works

  1. Read each source's saved position and login details.
  2. Pull changes since that position.
  3. Add or update sources and passages, and refresh who may read them.
  4. Mark anything the source has deleted or locked as removed.
  5. Record a health status for the source and move the saved position forward.

Pipeline

flowchart LR
  A[Connector] --> B[Saved cursor]
  B --> C[Upstream changes]
  C --> D[Passages]
  C --> E[Tombstones]
  D --> F[Health row]
  E --> F