wf-store

wf-store preserves immutable source assets, supports mutable tags and builds retrieval representations. Everything uses its supported front API and adds wiki-specific interpretation above it.

An asset is not a page by itself. Its source can be JSON, text or binary data; Everything's tags and policy determine whether it is a page revision, template or media asset and who can read it.

Read Assets, components and renditions for the data model, then One corpus per environment for namespace boundaries. Ingestion and search readiness explains why committed content can precede searchable content.

Search families and tradeoffs introduces exact, lexical and semantic retrieval. Retirement, preservation and recovery distinguishes removing public eligibility from deleting source.

Physical search units and AWS function names are implementation details. They must not become the logical identity of a wiki page or the durable location encoded in its links.

In this section

• Assets, components and renditions — Keep original source separate from the text used for retrieval.
• One corpus per environment — A namespace boundary, not a folder for each content type.
• Ingestion and search readiness — Commit, preparation, admission and searchable publication are distinct milestones.
• Search families and their tradeoffs — Different retrieval evidence, composed without inventing another content index.
• Retirement, preservation and recovery — Remove an asset from normal interpretation without destroying its source.