This documents how migrations from conversation history into the Knowledge Site are actually performed – established in practice across Sprints 1-7 of the Knowledge Preservation Project.
Guiding principle
Preserve knowledge, not conversations. The objective is to retain project understanding, not to archive every discussion verbatim. A migration that faithfully copies conversational back-and-forth has not actually preserved anything useful; a migration that extracts the conclusion and the reasoning behind it has.
Method
- Review conversations individually – one source at a time, not blended together.
- Extract enduring knowledge – conclusions, decisions, structures, reasoning – not the discussion that produced them.
- Ignore transient discussion – back-and-forth, dead ends, and small talk are not migrated.
- Merge duplicate knowledge – if two sources describe the same thing under different names, treat them as one and note the naming discrepancy rather than creating parallel records.
- Prefer expanding existing articles over creating new ones – a new article is only created when the knowledge genuinely doesn’t belong anywhere that already exists.
- Preserve historical decisions as proper Decision records, with their reasoning, not as bullet points buried inside unrelated articles.
- Cross-link related knowledge using the relationship system, so a reader arriving at any one record can find the rest of the picture.
- Verify relationships after every migration – check for broken links (a relationship pointing at a deleted or wrong-type record) before considering a migration finished.
Known limitation
Articles cannot currently link directly to other articles through the relationship system – only through plain in-content hyperlinks. Cross-linking between articles in practice relies on shared Decision and Task records (which can link to articles) plus manual hyperlinks. This has been noted as a recurring gap worth addressing if article-to-article linking becomes a frequent need.