fix(container): update image ghcr.io/defilantech/charts/llmkube ( 0.9.17 → 0.9.18 )

[renovate]

This PR contains the following updates:

Package	Update	Change
ghcr.io/defilantech/charts/llmkube	patch	0.9.17 → 0.9.18

---

Release Notes

defilantech/LLMKube (ghcr.io/defilantech/charts/llmkube)

v0.9.18

Compare Source

Features

• controller: architecture-aware runtime placement (#​1544) (6694d0b)
• controller: speculativeDecoding.draftModel for companion-file drafters (#​1564) (a8526a8)
• controller: surface actual GPU offload in InferenceService status (#​1566) (a2335c6)
• draft-model speculative decoding on InferenceService (#​1533) (81881d6)
• foreman: add optional Agent.spec.maxConcurrentTasks to bound in-flight tasks (#​1546) (0afa9c9)
• foreman: flag issue references removed by a diff (#​1574) (0f2d30d)
• foreman: honour the target repository's PR template when opening a PR (#​1562) (5d09e07)
• foreman: mutation-check decision logic (revert the fix, require a test to fail) (#​1578) (07754f1)
• foreman: operator-emitted task outcome counters (#​1563) (2386ad8)
• foreman: pre-flight staleness detection for already-addressed issues (#​1577) (cc9d594)

Bug Fixes

• controller: AND the arch requirement into nodeAffinity instead of ORing a new term (#​1584) (4e5e5d4)
• controller: do not roll over in-flight work when idleness is undetermined (#​1543) (daba4e1)
• controller: expose --dri-render-gid through the chart and stop defaulting it to the video group (#​1586) (66029e7)
• controller: inject the DRI render GID into Vulkan pod supplementalGroups (#​1569) (79f8f08)
• controller: let draft-mtp accept an optional draftModelRef (#​1547) (98fd3a0)
• controller: let ModelTransferHealthy recover after a retried download succeeds (#​1539) (978f5ee)
• deps: bump the Go toolchain to 1.26.6 for two stdlib advisories (#​1534) (627ca93)
• foreman: do not let an unavailable diff disable the grounded-finding rail (#​1576) (812186b)
• foreman: do not reserve a FleetNode for Job-mode AgenticTasks (#​1558) (55a1434)
• foreman: honour Workload spec.repo for gate, push and PR, not just clone (#​1545) (d949e3f)
• foreman: let the reviewer author an uncapped PR body via extra.prBody (#​1575) (827cad8)
• foreman: map test files to their modules in scope-overlap, and dedupe issue file refs (#​1573) (fa02f25)
• foreman: mount the agent's own gate-cache PVC, and honour an empty name (#​1557) (b09c3ad)
• foreman: remap unsupported emptiness claims to ERROR instead of NO-GO (#​1565) (ba7e3f9)
• foreman: restore the coder grounding rail's MCP prefix after the #​1527 rename (#​1582) (1b289e6)
• foreman: stop Job-mode AgenticTasks orphaning their previous Job on re-dispatch (#​1556) (32d1d65)
• namespace MCP tools with __ so names match the OAI pattern (#​1527) (1cd05d7)
• omit the issue reference when a task has no issue number (#​1532) (f881ac3)
• stop discarding a coder's diff when submit_result omits a message (#​1529) (3cd01ac)

Documentation

• make spec.files discoverable from spec.source for multi-artifact models (#​1540) (25777da)

---

Configuration

📅 Schedule: (UTC)

• Branch creation
  • At any time (no schedule defined)
• Automerge
  • At any time (no schedule defined)

🚦 Automerge: Disabled by config. Please merge this manually once you are satisfied.

♻ Rebasing: Whenever PR is behind base branch, or you tick the rebase/retry checkbox.

🔕 Ignore: Close this PR and you won't be reminded about this update again.

---

• If you want to rebase/retry this PR, check this box

---

This PR was generated by Mend Renovate. View the repository job log.