Skip to content

feat(evidence): publish the Flow 1.32.0 comparison; mark 1.31.0 superseded - #294

Merged
abrichr merged 1 commit into
mainfrom
feat/publish-flow-1.32-evidence
Aug 22, 2026
Merged

feat(evidence): publish the Flow 1.32.0 comparison; mark 1.31.0 superseded#294
abrichr merged 1 commit into
mainfrom
feat/publish-flow-1.32-evidence

Conversation

@abrichr

@abrichr abrichr commented Aug 21, 2026

Copy link
Copy Markdown
Member

What changed

  • publishes a fresh Flow 1.32.0 comparison, independent replication, transaction-fault probe, and remote-lease contract probe
  • binds the campaigns to the exact published wheel d58dcf8b…, sdist c3125f16…, release-tagged source d1b054c7… (v1.32.0), and Evals commit b7d4fe84…
  • marks the prior Flow 1.31.0 evidence as superseded instead of changing or deleting it
  • updates the readiness document and the version-pinned manifest test for the new generation

Why

The daily evidence-freshness workflow has been red since Flow 1.32.0 replaced 1.31.0 on PyPI (runs 32342633608, 32457392921). Published numbers described an engine nobody serves. This follows the PR #286 pattern: bounded campaigns, exact artifact binding, documented reproduction.

Counted result

93 counted trials against Flow 1.32.0, run locally (synthetic MockMed, loopback only, headless Chromium, no model calls):

Campaign Trials Result
Comparison 27 Compiled replay correct 9/9 across clean/theme/rename; selector controls halt safely under drift
Independent replication 27 Same shape in a second complete run: compiled 9/9 correct
Transaction probe 30 Uncertain deliveries return RECONCILIATION_REQUIRED; no blind retry or replay dispatch
Remote lease safety 9 Absent-or-changed frame leases refuse actuation (HALTED_BEFORE_EFFECT)

Zero silent incorrect success, zero wrong actions, zero over-halts, zero model calls across all counted arms.

Every campaign states production_acceptance: false. This set is development evidence, not a Production lifecycle admission, and establishes nothing about hosted, customer-workflow, native-desktop, RDP, or Citrix acceptance.

Validation

  • live PyPI check passes: bound to current published release 1.32.0 with matching wheel digest
  • offline evidence binding passes
  • 37 passed in tests/test_published_evidence_freshness.py

…seded

The daily evidence-freshness workflow went red when Flow 1.32.0 replaced
1.31.0 on PyPI: published numbers described an engine nobody serves.

Publishes a fresh Flow 1.32.0 comparison, independent replication,
transaction-fault probe, and remote-lease contract probe, run against the
exact published wheel (d58dcf8b...) and release-tagged source
(d1b054c7...), from Evals commit b7d4fe8. 93 counted trials:

- comparison and replication: compiled replay correct in 18 of 18 trials;
  selector controls halt safely under drift; zero silent incorrect success,
  wrong actions, over-halts, or model calls
- transaction probe: uncertain deliveries return RECONCILIATION_REQUIRED with
  no blind retry or replay dispatch
- remote lease probe: absent-or-changed frame leases refuse actuation

Every campaign states production_acceptance: false. The registry marks the
Flow 1.31.0 set superseded instead of changing or deleting it.
@abrichr
abrichr merged commit 08f025f into main Aug 22, 2026
2 checks passed
@abrichr
abrichr deleted the feat/publish-flow-1.32-evidence branch August 22, 2026 03:16
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant