Stream-scan lands and Augment-from-DB is feature-complete — scan a pulse stream, badge the known, one-click the new
Spec step 8 ships as a mode of search-and-add: a per-stream Scan button on the org card fires the official-blog machinery with the stream URL as the authoritative index, dedups against content_items, and badges what the corpus already holds — proven live against The Aspen Institute's blog, flag-flip and all. All five phases of the spec are now implemented.
Michael StatonMusic Editoraugmented with Claude Code on Claude Fable 5Stream-scan lands — and the flow the spec drew is all there
Phase 5 of [[../context-v/specs/Augment-From-DB-Flow.md]], and with it the spec goes Shipped: all five phases implemented in one day against a Signed-Off spec, each phase its own plan, commit, and proof.
What shipped
The cheap trick that made this a half-day, not a week: runOfficialBlogPack already had the exact seam — curated_index_urls, where operator-curated index pages skip discovery entirely. A stream scan IS "the index is already curated: it's the stream URL." So organization.stream.scan is a thin wrapper: fire the blog machinery at the stream, then dedup via the new content.urls.check on record-surrealdb-resolver (social-search has no DB access by design — the cross-service NATS read follows the domains.ts → content-ingest precedent).
On the org card: AdditiveList grew an optional per-entry action, and the Pulse-streams list wires it as scan — one click dispatches a stream-carrying envelope through the same localStorage + event + navigate contract as every 🔍.
In search-and-add: a stream envelope flips the surface into scan mode — TermBar and palette give way to the stream URL + Re-scan; result rows carry an "in corpus" badge with the ➕ parked; fresh items one-click into org_corpus through the existing add path, and the flag flips on the next scan.
Social walls (LinkedIn/Facebook) ride the same call but stay experimental per the spec's non-goal — blog/RSS/newsroom is the dependable path, and it's the one proven.
Proof
Live, end-to-end, against the stream Phase 2's proof seeded: scan of aspeninstitute.org/blog/ → 8 real posts with publish dates (July 2026, fresh) · organization.corpus.add one of them · re-scan → already_known: 1 and the added item's flag flipped to true. Full sweep green: svelte-check 0/0 on both remotes (one caught-and-fixed prop-destructure miss in AdditiveList — svelte-check earning its keep), all builds, three service typechecks, Phase 1 proof re-run 7/7.
Where the flow stands
All five phases of [[../context-v/specs/Augment-From-DB-Flow.md]] are implemented and pushed as attempt(augment-from-db, …, step1–5). Remaining before calling the product (not the code) done: the operator browser walk-throughs each plan names — the six-flow front door, the org card's ➕s, the 🔍→pairing→add loop, add-person with automatic affiliation, and the scan badge — plus the spec's parked open questions (fire-log persistence, pack-template seed terms, pinned-deploy remote URLs) for a future pass.