Package v0.2.12→v0.2.124: manifest, actions, version graph

startos/versions/v0.2.55.ts

@@ -0,0 +1,13 @@
+import { VersionInfo } from '@start9labs/start-sdk'
+
+export const v_0_2_55 = VersionInfo.of({
+  version: '0.2.55:0',
+  releaseNotes: {
+    en_US:
+      'Phase 1 diagnostic for the "missing analysis windows" issue: per-job audit-row drill-down on the dashboard\'s Jobs tab. (1) New backend endpoint GET /admin/job/:id/details returns every audit row keyed to a single job_id (transcribe row + one row per analyze window), pre-sorted by timestamp ascending so the pipeline reads in execution order. Includes a per-job summary with transcribe status, truncated-chunk count, total analyze rows, failed-window count, and a "windows planned vs windows logged" delta that flags jobs where analyze code planned N windows but only M < N audit rows landed (signature of a worker crash or silent dropped window). (2) New "Inspect" 🔍 column on the Jobs table. Click the magnifier next to any row to drop in a sub-row beneath it showing the per-window detail: timestamp, pipeline, window index, status pill, duration, model, and the recorded error message. Sub-row is lazy-loaded — first click fetches /admin/job/:id/details, subsequent toggles use the cached result. (3) Operator can now answer "WHICH window failed and WHY" without shell access to /data/relay-calls.ndjson. No fix to the underlying gap problem ships in this version — this release is purely diagnostic so we can confirm hypothesis 1 (single window silently failed) vs hypothesis 2 (transcribe chunk truncation upstream) on real failed runs before writing the retry + empty-sections-detection code for v0.2.56.',
+  },
+  migrations: {
+    up: async ({ effects }) => {},
+    down: async ({ effects }) => {},
+  },
+})