Anthropic only received a cache breakpoint on the last tool, so the large
frozen system prompt was never cached and multi-turn Patrol investigations
repaid its full input cost every turn (#2118). Serialize the stable base plus
mode text as a system content block with its own ephemeral breakpoint, leaving
the per-turn clock and accumulated knowledge outside the cached prefix.
The system breakpoint covers the tool definitions too, so the redundant tool
breakpoint is dropped when it is present. Providers without prompt caching
ignore the new request field and send the system prompt unchanged. Cache
creation/read counters are logged at Info so effectiveness is visible in
production.
Change-source: pulse-maintainer
Address the remaining verified findings from the pkg/metrics/store.go review
(#2079 finding 5 and the lower-impact nits):
- writeBatch now retries the whole transaction on a retryable lock error.
BEGIN is deferred, so contention is hit at the first INSERT or COMMIT; the
old loop retried only Begin() and skipped rows. While the one-time startup
maintenance (deferred identity-index rebuild and auto-vacuum VACUUM) is
active, an extended retry budget keeps upgrade-time history instead of
dropping it after the steady-state budget.
- migrateLegacyHostResourceType probes with an indexed EXISTS before opening a
write transaction, so an already-migrated store does not dirty the WAL on
every boot.
- QueryAll caps the per-series MetricPoint preallocation so a fine requested
step over a long range cannot reserve hundreds of thousands of slots.
Regressions: TestStoreStartupMaintenanceSignalsWriteRetryBudget,
TestStoreWriteBatchExtendsRetryDuringStartupMaintenance,
TestEstimateQueryAllBatchSeriesCapacityCapsPreallocation. The
performance-and-scalability contract records the invariants.
Change-source: pulse-maintainer
Distinguish independent customer-owned Community, Relay and Pro installations
from provider-owned MSP entitlements so outside IT support alone does not
trigger an MSP requirement. Preserve per-customer subscriptions and the
prohibition on pooling licences across unrelated customers.
Harden pkg/metrics/store.go rollup checkpoints, SQLite busy/locked retry classification, rollup min/max spread preservation, and one-connection auto-vacuum migration (#2079 findings 2-4, 6). Includes the performance-and-scalability contract sections, the regression proof moved into the registry-accepted store_additional_test.go, and the reviewed completion-pair registration for the split proof commit.
Change-source: pulse-maintainer
The canonical per-commit guard requires pkg/metrics/store.go changes to carry
an accepted proof file in the same commit. The proof for the rollup, retry,
upsert and auto-vacuum repair landed in the follow-up 087bbf8c20
(pkg/metrics/store_additional_test.go). Register the runtime commit e1d97b361e
and its completion commit as one reviewed completion unit so the candidate
preflight validates the pair instead of rewriting history.
Change-source: pulse-maintainer
The canonical completion guard requires the metrics-store hot-path proof in
pkg/metrics/store_additional_test.go. Move the rollup-gap, upsert-spread,
busy-retry and auto-vacuum regressions there and point the
performance-and-scalability contract at that file. The runtime change and
contract section are in the preceding fix(metrics) commit.
Change-source: pulse-maintainer
Address the remaining verified findings from the pkg/metrics/store.go review
(#2079 findings 2-4 and 6):
- rollupTier no longer leaps its checkpoint over a source gap. Previously a
trailing or interspersed gap advanced the checkpoint to the rollup cutoff, so
a lower tier that backfilled the gap later had those rows stranded below the
upper checkpoint and purged by retention.
- Retry conditions now classify SQLITE_BUSY/LOCKED by driver result code. The
modernc driver reports "database is locked (5) (SQLITE_BUSY)", which the
previous exact-string comparison never matched, so a busy writer dropped its
batch without retrying.
- The metrics upsert preserves an existing rollup row's min/max spread with
COALESCE instead of overwriting it with NULL.
- migrateAutoVacuum pins the pragma and VACUUM to one connection and verifies
the persisted mode, so the one-time conversion cannot silently repeat a
full-file VACUUM on every restart.
Regressions: TestStoreRollupTierBackfillsInterspersedGap,
TestStoreWriteBatchPreservesRollupSpread, TestIsRetryableWriteError,
TestIsRetryableWriteErrorMatchesDriverBusyCode,
TestStoreAutoVacuumPersistsAcrossRestart. The performance-and-scalability
contract records the invariants.
Change-source: pulse-maintainer
Restore the documented local-contention skip for load/SLO latency budgets on the shared release preflight worker: only a real GitHub-hosted run (GITHUB_ACTIONS=true and GITHUB_RUN_ID set) enforces the budgets. Hosted CI enforcement is unchanged; the worker keeps its cross-repo skip. Includes a focused predicate test.
Change-source: pulse-maintainer
Stop the physical-disk SMART merge letting a linked host agent's endurance reading raise a disk's Proxmox-reported remaining life, which resolved a genuine low-life disk-wearout alert and let it re-fire with a fresh resolved notification every poll interval (#2112). Includes focused regressions and the monitoring/agent-lifecycle contract sections.
Change-source: pulse-maintainer
Follow-up to the physical-disk endurance precedence fix. Records the
dependent agent-lifecycle contract section and moves the regression into the
accepted physical-disk proof file so later per-commit canonical guards find
both the contract and the verification artifact.
Change-source: pulse-maintainer
The physical-disk SMART merge let a linked host agent raise a disk's
remaining life above the Proxmox inventory value. An intermittent NVMe
PercentageUsed 0 reading therefore resolved a genuine low-life disk-wearout
alert, which re-fired on the next physical-disk poll and emitted a fresh
resolved notification every poll interval (#2112).
The merge now takes the agent value only when the Proxmox value is unreported
or more pessimistic, never raising a reported remaining-life value, while
still recording the agent SMART attributes. This does not change agent
admission, token binding or removal-block behaviour.
Regression: TestMergeHostAgentSMARTIntoDisks_AgentWearoutDoesNotHideLowPVELife
and TestMergeHostAgentSMARTIntoDisks_AgentWearoutFillsUnreportedPVELife in
internal/monitoring/monitor_integration_test.go. The monitoring contract is
updated in this commit.
Change-source: pulse-maintainer
Contract-Neutral: Behavioral bug fix in the physical-disk SMART merge; no public API or external contract delta. The monitoring contract is updated in this commit and the dependent agent-lifecycle contract and accepted proof file follow in the next commit.
The shared release preflight worker exports GITHUB_ACTIONS=true so isolated single-repository checkouts take their private-sibling skip. That flag also made the load/SLO latency guards hard-fail on a contended shared host, where the tests' own documented local-contention skip is the intended behaviour. Distinguish a real GitHub Actions run by GITHUB_RUN_ID, which only a hosted runner sets, so the worker keeps the cross-repo skip while latency overruns skip instead of failing. Hosted CI enforcement is unchanged.
Change-source: pulse-maintainer
Integrate the four user-facing core repairs: drop replayed unified metric samples (#1966), treat all local RepoDigests as the current image (#2110), honour the removal block on forked re-enroll (#2113), and stop closing writeCh on metrics-store shutdown (#2079 finding 1), with focused regressions and contract sections.
Change-source: pulse-maintainer
Restore the remaining focused regressions for the all-local-RepoDigests update comparison carried on the earlier core-runtime candidate: multi-digest resolution in getImageRepoDigests, false-update suppression in CheckImageUpdate, and the DigestsDiffer multi-digest cases. Test-only additions; the runtime change and contract sections are in the preceding fix(docker) commit.
Change-source: pulse-maintainer
The ingestion worker closed writeCh when it observed stopCh. A concurrent WriteWithTier that passed its stopping check, or a WriteBatchSync/WriteBatchBounded caller (which never checks stopping), could then send on the closed channel and panic the process during shutdown. Reproduced deterministically: enqueueWrite after Close panics with send on closed channel, and 8 concurrent monitoring writers racing Close panic in boundedEnqueueAndWait (store.go:1034).
Never close writeCh. Drain already-queued requests non-blockingly, then process them. Late writes land in the buffered channel and are discarded with the store rather than crashing it. Record the invariant in the performance-and-scalability contract and pin it in the accepted store proof file.
Change-source: pulse-maintainer
A removed host agent could re-enroll under a derived base-hex identity because identity resolution forked the machine before the removal block was consulted. The block is keyed on the base machine identity, so the forked ID bypassed it and the machine was silently re-admitted (#2113). ApplyHostReport now consults the base-identity removal block when the presenting token had no prior binding, so a fresh install token clears the block and heals to the base identity while an established distinct forked host (#1753) is left alone.
Record the behaviour in the monitoring and agent-lifecycle contracts and add the focused regression to the accepted monitor.go proof file. Change-source: pulse-maintainer
Change-source: pulse-maintainer
Docker update check compared a single RepoDigest against the registry tag digest, so a current image that carries two RepoDigests (postgres:16.15-alpine3.24) reported update-available. Compare the whole local RepoDigest set in CheckImageUpdate and in the typed update preflight; the primary digest is still reported for display.
Record the behaviour in the monitoring and agent-lifecycle contracts and add focused regression tests in the accepted docker proof files. Proof: go test ./internal/dockeragent/ passes including TestRegistryChecker_MultipleLocalRepoDigestsSuppressFalseUpdate and TestAgent_getImageRepoDigests_MultipleDigestsForOneImage.
Change-source: pulse-maintainer
Read-side registry rebuilds re-run the unified metric sync as often as every two seconds, so one poll observation was re-written many times between polls. The metrics store upserts on (resource, metric, tier, timestamp), so each replay UPDATEs the same row and commits a fresh transaction for no new data, churning the SQLite WAL and driving the disk-write amplification reported in #1966.
Track the last sample handed to the store per series and skip exact timestamp+value repeats before the batch is enqueued. A corrected value at the same observation time is still written, so the guard cannot mask a real change. Record the behaviour in the monitoring and agent-lifecycle contracts and move the focused regression into the accepted monitor.go proof file.
Change-source: pulse-maintainer
The canonical completion history validator materialised the completion tree with 'git worktree add', which writes administrative files into the reviewed repository's git directory. That directory is read-only inside the maintainer proof sandbox, so the maintainer preflight could not consult the registry that canonical governance already honours. Materialise the tree with a shared local clone instead: it reads the reviewed objects through alternates and writes only inside the disposable temporary directory, so CI and the maintainer pipeline validate the same reviewed pair without relaxing the sandbox boundary.
Change-source: pulse-maintainer
The Alerts overview stats table right-aligned the value cell, but a row carrying the small critical annotation rendered it inline after the count, pushing that count left of the other rows. Give the annotation its own right-aligned cell so the numeric column lines up, and assert the count cell stays numeric in the 24h regression.
Contract-Neutral: Behavioral alignment fix; no public-contract or subsystem-shape delta.
Change-source: pulse-maintainer
release_preflight_test still asserted the retired PVE runner labels after 051ce81a72 moved every release channel to GitHub-hosted runners. Assert runs-on: ubuntu-24.04 and the absence of self-hosted/pulse-pve so the test matches the governing workflow contract.
Change-source: pulse-maintainer
Initialize the persistent CSRF store for middleware tests that construct session-cookie requests, and restore the package test store afterward. This keeps exact release preflight runs from panicking before they can report qualification results.
Update the Pulse Account portal Vitest dependency to 4.1.11, clearing Dependabot security alerts 154 and 155 for Vitest and @vitest/mocker.
Contract-Neutral: Dev-only Vitest security dependency update; no product contract or runtime behavior changed.
Reloading configuration replaced the monitor's config pointer while API and
authentication handlers retained the original object. Tokens created afterwards
could authenticate successfully while Agent Doctor reported an unlisted token.
Refresh the canonical runtime configuration in place before constructing the
replacement manager. Cover repeated reloads, live token creation and revocation.
Use an exact runtime-file policy before the general API fallback, so setup preservation evidence cannot substitute for unrelated API changes. Assert both setup routing and unchanged organization/RBAC verification requirements. Retain the original failed broader-registry test evidence.
Change-source: pulse-maintainer
Sort the dedicated regression path into the existing exact-file list as required by the registry audit. Accepted proofs and enforcement remain identical.
Change-source: pulse-maintainer
The per-commit API guard needs the actual dedicated preservation test declared in its verification registry. Restore that tested artifact and add its exact path to the existing backend API proof list. Retain all existing requirements and reviewed commit identities; no history rewrite, contract-neutral override, or unrelated proof substitution is used.
Change-source: pulse-maintainer
Represent the retained L1 maintenance residual with the required kind and id fields so the status audit resolves its owned follow-up.
Change-source: pulse-maintainer
Link the L1 residual to an explicit owned follow-up when claiming the independent settings repair. Source admission, failed checks, installed recovery and exact-candidate observation remain unfinished rather than disappearing with the claim change.
Change-source: pulse-maintainer
Place the authenticated settings-preservation regression in the established API contract proof surface so the owning contract guard consumes its actual assertions, without waiving verification or changing runtime behavior.
Change-source: pulse-maintainer
Quick security setup previously replaced system.json with defaults even for authenticated force reconfiguration. Initialize only absent settings under the persistence save mutex, preserving existing bytes and read failures. Cover credential rotation, custom and malformed settings, missing settings, and read errors without changing authorization or token scopes.
Change-source: pulse-maintainer
Apply existing candidate content-drift validation to every future stable promotion, including v6.4 patches. Preserve patch and minor soak durations and the explicit hotfix reason path. Extend regression cases to maintenance releases and give historical unit fixtures explicit source paths rather than reading the live checkout.
Change-source: pulse-maintainer
Replacement PR checks were queued behind obsolete runs even after the
previous revision failed. Cancel prior validation only for pull_request
events, preserving completed verdicts for branch pushes and manual runs.
Contract-Neutral: Only PR concurrency changes. Semantic YAML comparison confirms all jobs and triggers, including frontend dependency security checks, are unchanged. No security verification change is warranted. The scheduling contract is updated.
Resolve the PR2087 frontend formatting gate without changing severity behavior or rewriting the reviewed source. Re-run the production-component browser matrix and refresh its content hashes; whole-tree formatting and local TypeScript checking pass.
Change-source: pulse-maintainer
pulse-triage[bot]