Skip to content

v1.5.1: harden cross-provider installation and hosted adoption #1050

Closed

Description

@jeffhuber

Goal

Ship a focused v1.5.1 reliability release from the successful v1.5.0 adoption campaign. No new product surface or Slack telemetry is in scope.

Five independent adoption passes—Claude, Cursor Cloud Agent, hosted Devin, a headless remote orchestrator, and Antigravity—installed or upgraded v1.5.0 without a release-blocking product failure. Their actionable installation, test-isolation, and documentation findings became the five roadmap revisions below.

PR map

Release boundary

Basic Slack lifecycle remains supported as released in v1.5.0 and requalified for v1.5.1. Slack-specific telemetry and hosted aggregation are deferred to v1.6.0 in #921 and #978.

Acceptance

  • Every implementation PR has focused regression coverage, exact-head independent review, green CI, and authoritative Code Mower gate.
  • Fresh hosted installation without uv/pipx, v1.5.0 upgrade, remote observer, safe init, Board, Graphify, and basic Slack rehearsals pass.
  • Single- and multi-lane audit publication succeed without owner override.
  • The bounded Board canary records provider exit, Board projection, privacy boundary, authorized usage, and settled usage separately.
  • Approved metadata-only evidence is uploaded to codemower.com and independently observed on a fresh authenticated dashboard view.
  • GitHub and PyPI v1.5.1 identities and artifact digests match the accepted release commit.

Activity

  1. jeffhuber commented on Sep 20, 2026

    @jeffhuber
    ContributorAuthor

    Documentation slice merged in #1051.

    Delivered from the four-provider v1.5.0 adoption evidence:

    • isolated, no-pipe uv bootstrap that works with externally managed system Python;
    • remote-only doctor --easy --orchestrator-only guidance;
    • local/path-bearing doctor JSON sharing boundary;
    • explicit role admission behavior for --dry-run and --no-lease;
    • explicit doctor-before-login order for headless Codex file authentication.

    Validation included an isolated public-package bootstrap, 363 release-hygiene tests plus 945 subtests, full Python 3.12/3.13/3.14 CI, Board and containment jobs, exact-head independent Claude PASS, and the authoritative gate. The workflow publisher's fresh source-seal refusal is recorded on #1032 rather than hidden by this merge.

    The behavior slices in this issue remain open: remote-observer doctor semantics, init/generated-output safety, and status/privacy polish.

  2. jeffhuber commented on Sep 20, 2026

    @jeffhuber
    ContributorAuthor

    Antigravity adoption evidence has been reconciled and the actionable test-harness slice is complete in #1052 (merged as ba72ac986ab3d24e1c5ace2fbcdcc9e4aed97aaa).

    Verified and incorporated:

    • pipx upgrade/install and core v1.5.0 CLI probes succeeded in the hosted Antigravity environment;
    • release-identity workflow subprocesses now put the active pytest interpreter first on PATH, so shell snippets invoking python cannot silently select an older ambient runtime;
    • the test suite now sets GIT_TERMINAL_PROMPT=0, so Git subprocesses fail deterministically instead of blocking on an unavailable credential prompt;
    • exact-head validation passed: 20 release-identity tests plus 116 subtests under a deliberately hostile ambient PATH, 363 release-hygiene tests plus 945 subtests, Ruff, diff check, full CI, and the Code Mower gate;
    • independent exact-head Claude audit: PASS, published by github-actions at Tests: isolate Python and Git subprocesses #1052 (comment).

    Corrections applied while curating the report:

    • the peeled v1.5.0 release commit is 197c0c6080afe2a7625d22b897f4027112cf4a01; the report's longer 197c0c609... value is not a valid release commit;
    • code-mower campaign run --repo ... --pilot is not a v1.5.0 CLI command and is excluded from guidance;
    • claimed full-suite, production-campaign, managed-persistence, and zero-downtime conclusions were not promoted because the supplied transcript did not establish them cleanly.

    This closes the Antigravity-specific deterministic local-test work. Broader cross-provider adoption work remains tracked by this epic.

  3. added this to the v1.5.1 milestone on Sep 20, 2026
  4. jeffhuber commented on Sep 20, 2026

    @jeffhuber
    ContributorAuthor

    v1.5.1 kickoff evidence was exported from exact clean main ba72ac986ab3d24e1c5ace2fbcdcc9e4aed97aaa and uploaded to codemower.com through the stored codex-code-mower profile.

    • Privacy boundary: metadata-only board_snapshot; 1 event, 0 reports.
    • Excluded: source, diffs, model transcripts, stdout/stderr, auth probe output, and secrets.
    • Health/token/bundle/model-provenance checks: PASS.
    • Stored receipt: HTTP 200, accepted 1, duplicate 0, upload ID 9312736d-4ccb-4ee4-b7e8-7a9fd8274252.
    • The receipt reported aggregate refresh stale; this proves storage only. A fresh authenticated dashboard observation remains required separately in Release v1.5.1: qualify, upload evidence, and publish #1056.
  5. jeffhuber commented on Sep 20, 2026

    @jeffhuber
    ContributorAuthor

    The v1.5.1 roadmap is complete.

    All five reliability revisions are merged, the hosted Board canary and basic Slack lifecycle are closed, exact-head qualification passed, the metadata-only release snapshot is stored on codemower.com, and the same immutable candidate is now published on GitHub and PyPI.

    Final release evidence is recorded in #1056. Slack telemetry and hosted aggregation remain in the v1.6.0 milestone as #921 and #978.

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Metadata

Assignees

No one assigned

    Labels

    epicEpic tracking issue

    Type

    No type

    Projects

    No projects

      Milestone

      Relationships

      None yet

      Development

      No branches or pull requests

      Issue actions