Commits · e79c4932c471005dcfc512ddfb3af28ac061e458 · OpenDAS / dynamo · GitLab

22 Apr, 2026 24 commits
- feat: Add vLLM start/stop profile endpoints (#8068) · e79c4932
  leo-cf-tian authored Apr 22, 2026
```
Signed-off-by: Leo Tian <lctian@nvidia.com>
```
  e79c4932
- fix(vllm): initialize failover flock in prefill worker (#8367) · 81819fbc
  mohammedabdulwahhab authored Apr 22, 2026
```
Signed-off-by: mohammedabdulwahhab <furkhan324@berkeley.edu>
```
  81819fbc
- feat(vllm): multinode elastic EP support via Ray cluster + concurrent scale fix (#8216) · c4964f55
  Tzu-Ling Kan authored Apr 22, 2026
```
Signed-off-by: Tzu-Ling <tzulingk@nvidia.com>
```
  c4964f55
- ci: remove unused SGLang NIXL ref (#8472) · e041ccfc
  Alec authored Apr 22, 2026
  
  e041ccfc
- refactor(mocker): replace vllm block manager with kvbm-logical (#8451) · 36b4208e
  Yongming Ding authored Apr 22, 2026
```
Signed-off-by: Yongming Ding <yongmingd@nvidia.com>
Co-authored-by: Ryan Olson <rolson@nvidia.com>
```
  36b4208e
- fix(runtime): reset canary health check timer on request arrival and per streaming chunk (#8467) · 1dc0975b
  Thomas Montfort authored Apr 22, 2026
```
Signed-off-by: Thomas Montfort <tmontfort@nvidia.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  1dc0975b
- docs: update docs for 1.1.0 DGDR (#8511) · cf92be04
  hhzhang16 authored Apr 22, 2026
```
Signed-off-by: Hannah Zhang <hannahz@nvidia.com>
Signed-off-by: hhzhang16 <54051230+hhzhang16@users.noreply.github.com>
Co-authored-by: coderabbitai[bot] <136622811+coderabbitai[bot]@users.noreply.github.com>
```
  cf92be04
- feat(router): branch sharded kv indexer (#7859) · 73a9a53f
  Janelle Cai authored Apr 22, 2026
```
Signed-off-by: Hannah Zhang <hannahz@nvidia.com>
Signed-off-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: Hannah Zhang <hannahz@nvidia.com>
```
  73a9a53f
- fix(vllm): cap gpu_memory_utilization for encoder-only vision model load (#8466) · af32579e
  Qi Wang authored Apr 22, 2026
```
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  af32579e
- fix(planner): match MDC component field against backend default, not DGD key (#8489) · 2cd4288a
  Hongkuan Zhou authored Apr 22, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  2cd4288a
- feat: address MR comments from TRTLLM/GMS integration (#8255) · dfd3aa99
  hhzhang16 authored Apr 22, 2026
```
Signed-off-by: Hannah Zhang <hannahz@nvidia.com>
```
  dfd3aa99
- fix(operator): run GPU discovery when any hardware field is missing (#8267) · d07288c5
  Dr. Stefan Schimanski authored Apr 22, 2026
```
Signed-off-by: Dr. Stefan Schimanski <sschimanski@nvidia.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  d07288c5
- docs(agents): improve tool/reasoning parser docs (#8497) · bae41d44
  Keiven C authored Apr 22, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  bae41d44
- fix: DIS-1768 emit structured SSE error + [DONE] on mid-stream fault (#8430) · f53fa64c
  Keiven C authored Apr 22, 2026
```
Signed-off-by: Keiven Chang <keivenchang@users.noreply.github.com>
```
  f53fa64c
- refactor(tests): consolidate diffusion payload classes into shared payloads.py (#8435) · 9bf9709a
  Krishnan Prashanth authored Apr 22, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
```
  9bf9709a
- fix(kvbm): Fix consolidator e2e tests (#8464) · d932d3b4
  Kris Hung authored Apr 22, 2026
  
  d932d3b4
- fix(sglang): populate total_kv_blocks via canonical Engine scheduler path (#8439) · 34acb1f4
  ishandhanani authored Apr 22, 2026
  
  34acb1f4
- feat: garbage collection tcp machine + read from shared place (#8420) · ffa6e84a
  Michael Feil authored Apr 21, 2026
```
Signed-off-by: michaelfeil <63565275+michaelfeil@users.noreply.github.com>
```
  ffa6e84a
- chore(mocker): fold workload concurrency cap into WorkloadDriver (#8446) · 73e0a2f2
  Yan Ru Pei authored Apr 21, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  73e0a2f2
- feat(kv-router): warn on duplicate store replays (#8473) · 4833e29b
  Yan Ru Pei authored Apr 21, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  4833e29b
- fix(profiler): pass total_gpus to AIC enumerate so candidates respect budget (#8479) · e9db304c
  Hongkuan Zhou authored Apr 21, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  e9db304c
- docs(observability): fix DYN_SYSTEM_PORT default in health-checks table (#8478) · d250fab9
  Jie Hao authored Apr 21, 2026
  
  d250fab9
- docs(profiler): clarify when planner-profile-data ConfigMap is emitted [DYN-2751] (#8486) · 5c35805e
  Hongkuan Zhou authored Apr 21, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  5c35805e
- fix(planner): DYN-2776 multi-DGD + GlobalPlanner scaling and readiness (#8482) · 6d44b904
  Hongkuan Zhou authored Apr 21, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  6d44b904
21 Apr, 2026 16 commits
- fix: Delete SQL tool call test for now since small models could occasionally... · 77a0051a
  Richard Huo authored Apr 21, 2026
```
fix: Delete SQL tool call test for now since small models could occasionally generate malformed arguments with guided decoding (#8474)
```
  77a0051a
- feat(global-router): add aggregated mode with TTFT x ITL pool selection (#8044) · 8af634ba
  jthomson04 authored Apr 21, 2026
```
Signed-off-by: jthomson04 <jwillthomson19@gmail.com>
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  8af634ba
- fix(discovery): send ModelUpdate::Added after add_worker_set completes (#8242) · 7ffc699f
  MatejKosec authored Apr 21, 2026
```
Signed-off-by: Matej Kosec <mkosec@nvidia.com>
```
  7ffc699f
- feat(lpu): add opaque engine_data field to nvext response with per-request opt-in via extra_fields · 1c9412d2
  Biswa Panda authored Apr 21, 2026
  
  1c9412d2
- fix(vllm): classify disagg decode-engine queued requests correctly in FPM (#8471) · 3f83c597
  Hongkuan Zhou authored Apr 21, 2026
```
Signed-off-by: Hongkuan Zhou <hongkuanz@nvidia.com>
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  3f83c597
- chore: rm kimi kvbm recipe | add qwen3 kvbm recipe (#8475) · d4ac0b59
  Ziqi Fan authored Apr 21, 2026
```
Signed-off-by: Ziqi Fan <ziqif@nvidia.com>
```
  d4ac0b59
- feat(profiler): wire mocker-rapid to direct AIC flags, drop profiler AIC interp (#8455) · 55a949cb
  Hongkuan Zhou authored Apr 21, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  55a949cb
- test: stabilize nightly — skip engine-init failures, convert xfails to skips,... · cc583b2f
  Dmitry Tokarev authored Apr 21, 2026
```
test: stabilize nightly — skip engine-init failures, convert xfails to skips, fix http URL validation regression (#8443)
Signed-off-by: Dmitry Tokarev <dtokarev@nvidia.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  cc583b2f
- chore: upgrade modelexpress version (#8336) · 9514236c
  Nate Mailhot authored Apr 21, 2026
```
Signed-off-by: Nate Mailhot <nmailhot@nvidia.com>
Co-authored-by: Dmitry Tokarev <dtokarev@nvidia.com>
```
  9514236c
- test: skip test_router_decisions_sglang_multiple_workers (DYN-2784) (#8441) · ddd19a68
  Tushar Sharma authored Apr 21, 2026
```
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  ddd19a68
- chore: copy attribution and license into planner image (#8453) · eb6c65d5
  Tushar Sharma authored Apr 21, 2026
```
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  eb6c65d5
- docs: document pipeline time budgets for lifecycle markers (#8374) · df888f68
  Tushar Sharma authored Apr 21, 2026
```
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  df888f68
- feat(tests): add --models-dir flag for read-only HF cache (#8362) · b6885977
  Ran Rubin authored Apr 21, 2026
```
Signed-off-by: rrubin <rrubin@nvidia.com>
```
  b6885977
- feat(kv-router): add start positions to flat KV events (#8426) · 4b3cd459
  Yan Ru Pei authored Apr 21, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  4b3cd459
- fix(planner): use dynamo-planner image for profiler job and planner pods... · e8ecf6ff
  Hongkuan Zhou authored Apr 21, 2026
```
fix(planner): use dynamo-planner image for profiler job and planner pods [DYN-2733][DYN-2746] (#8407)
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
Co-authored-by: Claude Opus 4.7 (1M context) <noreply@anthropic.com>
```
  e8ecf6ff
- fix: replace torch.load with safetensors and enable Rust frontend media... · f923777e
  Indrajit Bhosale authored Apr 21, 2026
```
fix: replace torch.load with safetensors and enable Rust frontend media decoding for TRT-LLM multimodal (#8295)
Signed-off-by: Indrajit Bhosale <iamindrajitb@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  f923777e