corpora-builder is born
Corpus building keeps reappearing inside every ai-labs product — so it gets a repo of its own, scaffolded with context-v, changelog, and a splash slot.
Michael StatonMusic Editoraugmented with Claude Code on Claude Fable 5Why Care?
Every deck, memo, enriched record set, and agent answer we ship is only as good as the corpus behind it — and corpus assembly has now shown up as a sub-feature in at least three products: augment-it's corpus inbox and funder corpora, memopop's research inputs, and dididecks' substantiation corpora. When the same shape recurs that many times, the tree's answer is to give it a home of its own. corpora-builder is that home: a project whose whole job is capturing, triaging, fetching, filing, and quality-checking source material.
What's New?
The repo exists:
lossless-group/corpora-builder, mounted as a submodule ofai-labs.Universal directories scaffolded per the pseudomonorepos convention:
context-v/(all six subdirectories) andchangelog/as root siblings.A
splash/slot reserved permaintain-splash-pages— known follow-up: the Astro splash itself is not yet scaffolded; Phase 1 ships without it.First design work underway:
context-v/explorations/Corpora-Builder-System-Design.md, which surveys the prior art (augment-it's Build-Corpora flow, the source-curation gate blueprint, the context-vigilance Chroma ingest) before any code gets written.
What's Next?
The exploration hardens into a spec once the option space settles — domain model, capture surfaces, curation gate, and the consumption contracts (Chroma ingest, deck/memo grounding). Patterns will travel to sibling apps knots-style — blueprint plus copy-from sample code — never as a shared package.