VCO Release v3.0.4

• Date: 2026-04-18
• Commit(base): c0618d0
• Previous public release: v3.0.3
• Previous public release commit: 48743c5

Highlights

• Promoted the latest maintained main state after v3.0.3 instead of leaving the direct-routing and canonical-entry hardening work stranded behind the public line. This cut packages the current maintained source at c0618d0.
• Replaced the odd "open a Codex child session to ask a specialist" behavior with direct in-session specialist routing. The governed runtime now routes specialists to the current session by default, keeps the accounting and proof surfaces explicit, and preserves reviewable fallback truth instead of hiding specialist decisions behind extra host-specific conversation launches.
• Hardened canonical vibe entry truth end to end. Runtime packet coverage, host launch receipts, launch-failure receipt state, and canonical truth-artifact checks now line up more tightly with what the governed runtime actually emits.
• Repaired a release-train contradiction around adaptive-routing evidence. The adaptive readiness gate now self-seeds route telemetry through the observability gate when a clean checkout has no prior outputs/telemetry evidence, so governed release cuts no longer depend on forbidden tracked outputs or a manually primed working tree.
• Tightened upgrade correctness around installed-target truth and request defaults. Target-root upgrade checks now read the right truth source, specialist snapshot-ignore handling no longer misclassifies consultation state, and empty /vibe-upgrade invocations now default to the shared upgrade path for the current host instead of stopping to ask the user what should be upgraded.

What Changed Since v3.0.3

• v3.0.3 moved the public line forward to the then-latest maintained source at 537db3f, focusing on host-global bootstrap lifecycle, specialist decision truth, upgrade metadata hardening, and installed-runtime payload completeness.
• v3.0.4 continues that 3.0.x line with the next maintained source after the v3.0.3 cut:
  • PR #170 hardened canonical vibe runtime entry behavior and proof expectations.
  • PR #171 restored canonical entry runtime payload coverage so the shipped runtime carries the files the governed entry actually depends on.
  • PR #172 repaired native specialist Codex home seeding and sidecar cleanup so direct specialist execution could stabilize without host-specific bootstrapping regressions.
  • PR #174 moved specialist routing to the current session by default, followed by review fixes around accounting, recommendation-phase matching, and degraded consultation proof.
  • PR #175 fixed host-launch receipt lifecycle so launch failures mark the receipt as failed instead of leaving a misleading launched-only state behind.
  • PR #176 corrected target-install upgrade truth so upgrade verification reads the selected target root instead of drifting back to the wrong source.
  • PR #177 fixed specialist consultation snapshot-ignore handling for issue #173, closing a runtime truth bug around consultation artifacts.
  • PR #178 fixed empty /vibe-upgrade requests so they resolve to the shared upgrade intent for the current host instead of degenerating into an empty governed request.
• The release cut itself also closes a stop-ship gap that existed on latest main: adaptive-routing readiness no longer assumes pre-tracked telemetry under outputs/telemetry, and instead generates bounded replay evidence through the governed observability path during verification.
• The practical result is a more believable governed runtime surface: specialists route directly, canonical entry proof is stricter and more honest, release gating is reproducible from a clean checkout, and upgrade behavior is less likely to ask the user unnecessary clarifying questions when the intended action is already obvious.

Release Positioning

• This is a 3.0.x continuation release, not a new major baseline.
• The release is centered on behavior correctness and runtime truth rather than on opening a brand-new product lane. It closes gaps that were especially visible in real use: specialist routing semantics, canonical launch proof lifecycle, target-root upgrade verification, and empty upgrade invocation behavior.
• As with v3.0.3, package metadata remains on the repository's current 3.0.0 package line. The meaningful public version signal for this cut is the governed release surface at v3.0.4.

Validation Notes

• pwsh -NoProfile -ExecutionPolicy Bypass -File ./scripts/governance/release-cut.ps1 -Version 3.0.4 -Updated 2026-04-18 -Preview -RunGates -> preview succeeded.
• pwsh -NoProfile -ExecutionPolicy Bypass -File ./scripts/governance/release-cut.ps1 -Version 3.0.4 -Updated 2026-04-18 -RunGates -> release cut complete.
• pwsh -NoProfile -ExecutionPolicy Bypass -File ./scripts/verify/vibe-adaptive-routing-readiness-gate.ps1 -WriteArtifacts -> PASS from a clean state by auto-seeding replay telemetry through vibe-observability-gate.ps1.
• python3 -m pytest tests/integration/test_release_cut_gate_contract_cutover.py -> 3 passed.
• git diff --check -> PASS.

Migration Notes

• Operators upgrading from v3.0.3 should expect direct in-session specialist routing to be the normal path. The governed runtime should no longer need a host-specific Codex child conversation just to route a specialist during ordinary execution.
• If you rely on canonical-entry truth artifacts, this release is stricter about host-launch receipt lifecycle and runtime payload completeness. Failure states and proof expectations should now be easier to audit.
• If you use /vibe-upgrade or similar host-visible upgrade wrappers, this is the first public cut where an empty invocation defaults cleanly to the shared current-host upgrade path instead of collapsing into an empty governed request.