An editable knowledge base with stable reader copies

The first requirement was a secure internal website backed by Markdown. People needed to find documents, see where they came from and retain their edits after a service restart. The delivered site included navigation, search and edit history.

The next phase addressed how the same knowledge could remain available to file-based readers as the editable content moved into a database.

The database became the place to edit and store authoritative content. A publication process generated read-only Markdown snapshots from it, preserving a file-based route for discovery.

A document change belongs in the database. Readers receive a generated snapshot: a published copy with no separate editing path to reconcile later.

The publisher records the source revision before building a snapshot. It checks the revision again and rejects the build if the authoritative content changed during preparation. That build cannot be published as though it came from one unchanged revision.

The process verifies the snapshot, makes it read-only and publishes it. It then checks the published result and records the files, their content fingerprints and the publication run. These records let an operator trace what was made available to readers.

If publication fails after exposing a snapshot, the implementation rolls that publication back.

The editable website was delivered in July 2026. A later July release added database authority and generated reader copies.

The initial scope covered finding and maintaining documents. The expansion added a way to publish that knowledge for file-based readers, with revision checks and a record of each published snapshot.