ODS V3 Pre-Release: Public testing and refinement ahead of the official V3 launch. Turn your PC, Mac, or Linux box into a private AI server.
Extension catalog source and qualification status
Reviewed 2026-09-24 UTC against main promotion
1bc5e1cbb24864f1c9efd551e0613202af324e3f.
PR #6156 merged on 2026-09-23 at
1d6038ef9679b89632f5fa4765ad1d6347786711; its changes are included in the
September main promotion. The former
instruction to wait for PR #6155 is not a current merge dependency. PR #6155
remained an open draft when this record was refreshed; its state does not prove
or disprove the current tree's platform behavior.
Reproducible source evidence
| Area | Evidence at the reviewed source | Boundary |
|---|---|---|
| Generated catalog | config/extensions-catalog.json: 200 entries, 200 unique IDs |
Presence does not prove deployability or application quality |
| Extension definitions | python scripts/audit-extensions.py --project-dir . --include-library --strict: 203 services, zero errors/warnings |
Structural source audit; no application images pulled or started |
| Library source | 171 library directories, 137 upstream.json records |
34 recipes lack structured provenance; the audit does not clear their terms |
| Pixel source custody | python scripts/verify-pixel-bundle.py: pass |
Matching visible source and a single-commit installation bundle; not installed acceptance |
| Platform and model behavior | See the promotion record and Portal regression acceptance | No new physical-machine or live-model acceptance is asserted here |
Catalog requests use existing recipes; source-project preparation and managed installation have separate evidence boundaries. A proposal, successful import, prepared recipe, running process and healthy application are different states. Record which state an acceptance test establishes.
The earlier page included transient workstation deployment claims and test counts from multiple revisions. Those do not establish acceptance at the promoted commit. Future receipts must bind product SHA, command, platform, runtime identity, outcome and retained evidence. Do not publish private session IDs, personal paths, credentials or conversational content in this public page.
Remaining acceptance
Exercise actual application use, required configuration, failed install/repair, cancellation, reconnect, repeated requests, and platform-specific lifecycle recovery before describing the entire catalog as qualified. Static catalog and unit-test success do not supply that evidence. Current work must preserve the runtime testing campaign's selected versions and settings.