Add multi-tenant cloud mode: self-serve purchase, credit metering, core-decoupling

Introduces RECAP_MODE=multi alongside single-mode self-host:
- Tenant auth + accounts (magic-link via System SMTP), per-tenant credit pool,
  anonymous trial minting with per-IP/-64 caps
- Self-serve Pro/Max purchase: inline Lightning (BTCPay) + card (Zaprite),
  prepaid 30-day periods, expiry-reminder emails
- Core-decoupling: relay owns cloud tier/expiry keyed by Recaps user-id
- SQLite (better-sqlite3) schema for multi-mode; filesystem unchanged for single
- StartOS actions/versions through 0.2.155
This commit is contained in:
Keysat
2026-06-13 14:25:05 -05:00
parent db580abad7
commit 0ae59f3550
176 changed files with 23823 additions and 803 deletions
+13
View File
@@ -0,0 +1,13 @@
import { VersionInfo } from '@start9labs/start-sdk'
export const v_0_2_74 = VersionInfo.of({
version: '0.2.74:0',
releaseNotes: {
en_US:
'Recap-app side of the relay-mode speaker-identification fix (paired with relay 0.2.63). When /api/process is invoked in relay-mode (both transcribe-provider and analyze-provider set to "Recap Relay"), Recap-app now fetches YouTube metadata via fetchYouTubeMetadata() BEFORE handing the URL off to the relay — channel, description (capped at 2000 chars), and chapters — and passes all three through to /relay/summarize-url\'s request body as channel / description / chapters fields. Previously this branch sent `channelHint: ""`, `descriptionHint: ""`, `chaptersHint: []` even though the same metadata fetcher was available in scope — the relay was running blind on every Recap submission. The fetch costs ~1-2 seconds (negligible relative to the 1-3 minute summarize pipeline) and is gated on `!isPodcast && videoId` so podcast submissions skip it (fetchYouTubeMetadata is YouTube-specific). Combined with relay 0.2.63\'s own yt-dlp metadata extraction (which fills in any fields the client sends empty), every Recap → Relay submission of a YouTube URL now produces transcripts with proper speaker labels — "Brandon Karpeles:" and "Matt Hill:" rather than unlabeled dialogue. Both packages can be upgraded independently: an older relay paired with this Recap version still gets the speaker labels (Recap fills them in); a new relay paired with an older Recap version also gets them (relay\'s yt-dlp extraction fills them in).',
},
migrations: {
up: async ({ effects }) => {},
down: async ({ effects }) => {},
},
})