fix(container): update image ghcr.io/defilantech/charts/llmkube ( 0.9.15 → 0.9.17 )

[renovate]

This PR contains the following updates:

Package	Update	Change
ghcr.io/defilantech/charts/llmkube	patch	0.9.15 → 0.9.17

---

Release Notes

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

v0.9.17

Compare Source

Features

• chart: foreman.crs.namespace — render the CRS ConfigMap where kube-state-metrics can mount it (#​1492) (5bb8c19)
• fetch_pull_request: include check-run output (title/summary/text) for failed runs (#​1481) (#​1505) (e323f32)

Bug Fixes

• chart: heartbeat alert uses time(), not the nonexistent time_seconds() (#​1494) (bf8c1a0)
• chart: scope agent Deployment selectors to their pool (#​1524) (0da2a9e)
• detect the scheduler's other wording for an unusable cache volume (#​1513) (64ee9d7)
• EnsurePR treats a closed unmerged PR as satisfied — retried work ends with no PR (#​1476) (#​1507) (82baff8)
• extract path refs the extension gate was discarding (#​1516) (7166920)
• foreman: bound edit-free resets earned from bash-shaped writes (#​1521) (41a2829)
• foreman: bound grep match lines so one minified file cannot end a run (#​1488) (952f8af)
• foreman: one tool list, so a tool cannot ship unusable again (#​1486) (e77d7cf)
• foreman: reviewers review the change, not the base branch (+ cross-project remote guard) (#​1490) (0f3fa83)
• foreman: route all three consumers through the provider seams (#​1510) (79448fc)
• modelpool: guard the nil HTTPClient that panics every swap (#​1485) (dc3fa50)
• name the cause when a model cache PV can never bind (#​1511) (772c99d)
• normalizeContainers strips operator-owned SecurityContext fields before drift comparison (#​1462) (#​1506) (8d291ca)
• operator: apply nodeSelector and tolerations to non-GPU services (#​1504) (dc3bac3)
• populate status.size and status.gguf for s3:// model sources (#​1522) (69a76fd)
• router-proxy: honour 0 as disable, pin the metrics port in tests, refresh the dashboard catalog (#​1519) (774e0b9)
• router-proxy: reject colliding listener addresses at startup (#​1518) (fc49a5d)
• surface the unbindable-cache diagnosis instead of dropping it (#​1512) (03c0f17)

Documentation

• foreman: refresh model compatibility from what the fleet has proven (#​1483) (8ccdbee)
• how to recognise a cache volume the provisioner never stamped (#​1514) (12d59c2)
• operations: document ModelPool in the GPU-sharing guide (#​1508) (c2c3a84)

v0.9.16

Compare Source

Features

• chart: PodMonitor for router-proxy metrics (#​1473) (fd7c640)
• foreman: advisory two-site parity signal in the coder gate (#​1456) (944f90d)
• foreman: image input, so a vision-capable coder can see what it rendered (#​1471) (6c122be)
• foreman: read-only fetch_pull_request tool (#​1458) (7f4910b)
• modelpool: cross-pod sticky model swapping on a shared GPU slot (#​1393) (ba7777d)
• router-proxy: expose llmkube_router_* metrics on a dedicated listener (#​1457) (ca3730d)

Bug Fixes

• controller: make the custom CA additive to the system trust store (#​1469) (d490059)
• controller: stage files and mmproj from an s3:// object store (#​1475) (769b60d)
• dashboard: AMD dashboard vendor hygiene — remove runtime-agnostic llama.cpp panels, standardize on DS_PROMETHEUS (#​1446) (c6c0670)
• dashboard: retarget amd-gpu-observability off amdgpu_*-only queries (#​1474) (a02e8e8)

---

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.