Skip to content

clio-relay 1.4.4

Choose a tag to compare

View all tags
@JaimeCernuda JaimeCernuda released this 19 Jul 18:21
v1.4.4
cf713c9
This commit was created on GitHub.com and signed with GitHub’s verified signature.
GPG key ID: B5690EEEBB952194
Verified
Learn about vigilant mode.

This patch fixes a live availability defect exposed by long-running ParaView/JARVIS services: cluster-side status, progress, artifact, and runtime-binding queries now have reserved endpoint capacity instead of waiting behind the workload they need to inspect.

Highlights:

  • Persists a generic per-cluster worker-capacity policy (three total slots, one reserved control slot by default).
  • Derives control admission server-side from pinned JARVIS semantics or fresh, immutable, cluster-owned remote-MCP discovery evidence; callers cannot claim the lane themselves.
  • Bounds reserved queries, rejects stale or tampered evidence, and preserves legacy workload retry identities.
  • Makes worker status generation-safe and restart-only operations fail closed when the installed unit has drifted from policy.
  • Fails secure-runtime acceptance before scheduler submission when reserved capacity is unavailable.
  • Adds relay-only idempotency keys for safe retries of generalized virtual MCP tools.

Focused unit, integration, adversarial-security, type, lint, package, and release-policy checks passed locally. Released-wheel Ares validation follows immediately; no scheduler jobs are cancelled by default.

Assets 5
Loading