Issues

• Local shutdown reports success when the helper fails (exit status ignored)

  MiaAI-Lab/sparkDash#379 · Filter by author simonimhotep opened on Nov 8, 2026

• docker-compose does not forward MODEL_PRICES_FETCH / MODEL_PRICES_JSON_PATH — Cost compare cannot be enabled on a Docker install

  MiaAI-Lab/sparkDash#378 · Filter by author simonimhotep opened on Nov 8, 2026

• Overview: multi-rank node without HTTP listener only shows 'GPU is in use' — worker-role fix is undiscoverable

  MiaAI-Lab/sparkDash#373 · Filter by author micachi opened on Nov 8, 2026

• Commit e89f8d1 is missing file

  MiaAI-Lab/sparkDash#370 · Filter by author mapfde opened on Nov 6, 2026
  2

• Feature: Add the possibility to see TensorFold/vLLM version

  MiaAI-Lab/sparkDash#369 · Filter by author nemomaximus opened on Nov 6, 2026

• Fleet energy goes blank when one unit is offline

  MiaAI-Lab/sparkDash#359 · Filter by author markyu0401 opened on Nov 2, 2026
  3

• LLM probe: add absolute KV token counts (tile sub-line + history corner), populate them for the tensorfold backend

  MiaAI-Lab/sparkDash#348 · Filter by author de-flandres opened on Nov 1, 2026
  2

• Shared tiered range picker with custom spans, adopted everywhere a time window is chosen

  MiaAI-Lab/sparkDash#347 · Filter by author de-flandres opened on Nov 1, 2026
  1

• Compare view: render the same history panels for the whole fleet on one page/time axis

  MiaAI-Lab/sparkDash#345 · Filter by author de-flandres opened on Nov 1, 2026
  1

• Prefill bench: over-time rate view for a running long-context test + participant line for multi-rank engines

  MiaAI-Lab/sparkDash#344 · Filter by author de-flandres opened on Nov 1, 2026
  1

• KV cache panels: expose the per-pool breakdown in the tooltip (vector read of the worst-lane ratio)

  MiaAI-Lab/sparkDash#342 · Filter by author de-flandres opened on Nov 1, 2026
  1

• Fleet Energy: 'Efficiency' panel charts consumption-per-token (lower is better) under an efficiency name — rename and/or invert to tokens-per-Wh

  MiaAI-Lab/sparkDash#341 · Filter by author de-flandres opened on Nov 1, 2026
  1

• RoCE/RDMA utilization over time: persist per-device RX/TX + loss trends (data already collected)

  MiaAI-Lab/sparkDash#335 · Filter by author de-flandres opened on Nov 1, 2026
  2

• Storage usage over time: historize per-mount usage (worst-mount % / largest-mount used) and add a history chart

  MiaAI-Lab/sparkDash#334 · Filter by author de-flandres opened on Nov 1, 2026
  2

• Record + chart LLM series the probe already computes: avg decode tok/s per stream, cached/uncached prefill split, MTP acceptance

  MiaAI-Lab/sparkDash#333 · Filter by author de-flandres opened on Nov 1, 2026
  1

• Feature: running/queued lanes over time on the LLM history chart (see inflow pressure vs prefill/decode/power)

  MiaAI-Lab/sparkDash#331 · Filter by author de-flandres opened on Nov 1, 2026
  1

• Proposal: optional "API Keys" page for LiteLLM virtual keys (happy to contribute a PR)

  MiaAI-Lab/sparkDash#297 · Filter by author travellingsoldier85 opened on Oct 28, 2026
  4

• [Feature Request] Serve under a URL sub-path behind a reverse proxy (SPARKDASH_BASE_PATH)

  MiaAI-Lab/sparkDash#192 · Filter by author naeemakhtar110 opened on Oct 11, 2026
  11