promote: verified develop snapshot 89f673c618f to main (cold-start TTL + onboarding Bug B — final sync) - #15543
Merged
Merged
Conversation
fix(ci): make app aesthetic audit advisory
Co-authored-by: Shaw <shawgotbags@gmail.com>
Co-authored-by: Shaw <shawgotbags@gmail.com>
Co-authored-by: Shaw <shawgotbags@gmail.com>
…hetic-advisory # Conflicts: # packages/app/scripts/ios-device-e2e.mjs # packages/app/scripts/lib/ios-device-e2e-lib.mjs # packages/app/scripts/lib/ios-device-e2e-lib.test.mjs # plugins/plugin-openai/__tests__/streamstructured-tool-input-delta.shape.test.ts
fix(ci): generate i18n data before hetzner reaper
…hetic-advisory # Conflicts: # .github/workflows/hetzner-e2e-reaper.yml
… spike) Measured on prod f6f91af tonight: the first message after >60s idle pays preforward total=2959ms (auth=1684, mid=482, reserve=705) vs 109ms warm — a 3-5.5s first-turn spike whose dominant term is the IAC entry expiring between natural chat pauses. The IAC's primary freshness mechanism is explicit confirmed-delete invalidation on every credential mutation (revoke/update/ban/deactivate, all fail-closed per #13417); the TTL is only the backstop for an invalidation that was never issued. 300s keeps active conversations warm across gaps while bounding a lost-invalidation window to the same 5 minutes org.data already accepts.
…of dead-ending 503 during a provisioning-worker outage Native cloud onboarding provisioned+503'd for a user whose one healthy agent needed no provisioning at all (#15516 Bug B). Two stacked defects: - client: #15491 made the reuse pick running-only, so ANY transiently non-running row (starting/stopped/status churn during the very outage that 503s creates) fell through to a fresh create; the reuse branch's cold-boot wait had become unreachable dead code. The pick now prefers running, then falls back to the newest NON-TERMINAL row routed through waitForCloudAgentRunning (which kicks a resume and re-reads the fresh post-wake record — the correct guard against binding a stale pointer, preserving #15491's intent). A remembered preferAgentId is honored whenever non-terminal, so we never silently swap a user onto a different agent's conversations. - server: POST /api/v1/eliza/agents ran the provisioning-worker health gate BEFORE createAgent's idempotent reuse, so a heartbeat gap 503'd even callers whose create would have been served by reuse (no job enqueued). The route now peeks reuse candidacy (read-only repository mirror of the guard predicate) and only gates a create that would genuinely enqueue; forceCreate always stays gated. Repairs 3 pre-existing red tests in the mock-cloud connect suite (the running-only pick broke the stopped-agent reuse cases on develop) and adds coverage: remembered-agent wake, all-terminal create fall-through, and route-level gate-vs-reuse ordering (reuse-skips-gate, no-agent still 503s fail-closed, forceCreate never skips). Fixes #15516 (Bug B)
…auth-test fix(ci): make credential tunnel route test auth explicit
…lable fix(ci): skip hetzner reaper on rejected token
fix(ci): repair scenario-pr e2e harness resolution
perf(cloud): raise inference auth-context TTL 60s → 300s — kills the 3-5.5s first-message-after-idle spike
…hetic-advisory # Conflicts: # .github/workflows/hetzner-e2e.yml # packages/app-core/src/api/credential-tunnel-routes.test.ts
fix(cloud): reuse the existing agent in first-run onboarding instead of 503 dead-end during worker outage
NubsCarson
moved this from In progress
to Done
in Launch QA — Eliza app + Cloud (#13406)
Contributor
|
Claude encountered an error —— View job I'll analyze this and get back to you. |
Contributor
|
❌ PR title does not match the required pattern. Please use one of these formats:
|
Member
Author
[vps-backend] NIGHT CLOSEOUT — full sync achieved, everything shipped/tested/deployed ✅Prod ( Verified live on prod just now:
Ledger of the night (all merged + deployed): #15508 mid-reads ∥ (Shaw) · #15520 embeddings passthrough · #15521 placement staging-soak · #15524 reply-first streaming · #15526/#15525/#15532 onboarding cleanup ([codex-ui]) · #15541 Bug B both halves ([cloud-agent]; my #15542 closed as superseded — good race to lose) · #15538 cold-start TTL · #15511 pair CORS. Prod chat: 5.3s → ~1.0s warm, ~2.5s after-idle, streaming real. Open + owned (nothing unowned): #15539 [cloud-agent] · #15545 [codex-ui] · #15540 routed to CI/native lane (was unowned — needs an owner with NDK context) · #15513 placement prod-leg (mine, after staging soak) · #15518 decision posted (keep shared bridge; dedicated provision still ~40s). Team worked like a team tonight. 🫡 |
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
2 participants
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Promotes verified develop snapshot
89f673c618fto main — 55 checks, 0 failures on the tip (cleanest snapshot today). Final sync of tonight's launch-hardening arc; brings main even with develop.Carries (29 commits):
auth=1684mscold → warm across natural chat gaps)reuseWouldServeserver gate)After merge: CF Deploy main→production (direct-wrangler bypass ready if the scheduler wedges again per #15503) → prod probe: cold-start spike gone (first-after-idle ≤ ~2s), embeddings+chat passthrough intact, onboarding create-during-blip returns 200 reuse.
Rollbacks: all constituent fixes independently revertible (TTL constant, route-scoped, plugin-scoped).
🤖 Generated with Claude Code