Commits · 76c96c5d6c6ba499dd3af3a776c5cf830f02f076 · OpenDAS / dynamo · GitLab

06 Mar, 2026 12 commits
- chore(frontend): Remove the debug_perf flag used for perf work (#7024) · 76c96c5d
  Graham King authored Mar 06, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  76c96c5d
- feat(trtllm): add additional metrics for dynamo-trtllm (#6668) · 10081929
  Yifan Jiang authored Mar 06, 2026
```
Signed-off-by: Yifan Jiang <19356972+yifjiang@users.noreply.github.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  10081929
- chore: add mypy typing to vllm (#6858) · 2831bfec
  jh-nv authored Mar 06, 2026
  
  2831bfec
- chore(frontend): Remove the multi-processing vllm processor path (#7005) · 35b0ce62
  Graham King authored Mar 06, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  35b0ce62
- fix: guard SGLang/vLLM memory occupation control endpoints (#6967) · c09a9aad
  Schwinn Saereesitthipitak authored Mar 06, 2026
  
  c09a9aad
- fix: llm/mocker: Remove the llm -> mocker crate dependency, move config (#6998) · abc02c68
  Graham King authored Mar 06, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  abc02c68
- fix: restore --enforce-disagg to reject requests before prefill router activates (#6957) · f934744f
  Yan Ru Pei authored Mar 05, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  f934744f
- fix: TRT-LLM multimodal preprocessor - remove default_multimodal_input_loader... · e6ddf0ea
  Michal Guzek authored Mar 05, 2026
```
fix: TRT-LLM multimodal preprocessor - remove default_multimodal_input_loader from the embedding paths (#6924)
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
```
  e6ddf0ea
- fix: propagate tolerations and cap auto-discovered GPUs (#6947) · b97fde10
  hhzhang16 authored Mar 05, 2026
```
Signed-off-by: Hannah Zhang <hannahz@nvidia.com>
```
  b97fde10
- fix: efficiently fill dummy data to bypass vLLM preprocessor in E/P/D (#6968) · 503d042a
  GuanLuo authored Mar 05, 2026
```
Signed-off-by: Guan Luo <41310872+GuanLuo@users.noreply.github.com>
```
  503d042a
- chore: add mypy typing for trtllm (#6860) · f99b78f0
  jh-nv authored Mar 05, 2026
  
  f99b78f0
- fix: reject prompts exceeding max_seq_len with HTTP 400 (#6635) · 124ecd98
  Yuewei Na authored Mar 05, 2026
```
Signed-off-by: Yuewei Na <nv-yna@users.noreply.github.com>
Co-authored-by: Yuewei Na <nv-yna@users.noreply.github.com>
```
  124ecd98
05 Mar, 2026 9 commits
- chore: add mypy typing to sglang (#6859) · 6398d453
  jh-nv authored Mar 05, 2026
  
  6398d453
- fix(vllm-omni): Fix call to normalize_finish_reason on OmniHandler for main (#6910) · 22f1ab15
  Ayush Agarwal authored Mar 05, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
  22f1ab15
- fix: Enforce min_endpoint flag in Planner (#6637) · f1e8ea6e
  Karen Chung authored Mar 05, 2026
  
  f1e8ea6e
- fix: revert accidental change of trtllm/multimodal_processor (#6921) · 9c6f36e9
  Qi Wang authored Mar 05, 2026
```
Signed-off-by: Qi Wang <qiwa@nvidia.com>
```
  9c6f36e9
- chore: add mypy typing for components/src/dynamo/common (#6721) · 869e7336
  jh-nv authored Mar 04, 2026
  
  869e7336
- perf: multimodal benchmark sweep (#6795) · 9780bf3a
  Qi Wang authored Mar 04, 2026
  
  9780bf3a
- fix: Skip Encoder llm creation for unsupported models in trtllm (#6866) · f0bfda1e
  Indrajit Bhosale authored Mar 04, 2026
```
Signed-off-by: Indrajit Bhosale <iamindrajitb@gmail.com>
```
  f0bfda1e
- feat: Add NVTX markers for vLLM EPD (#6627) · eb0bf24e
  Kris Hung authored Mar 04, 2026
  
  eb0bf24e
- fix: revert changes in autogen files (#6908) · e15685d8
  jh-nv authored Mar 04, 2026
  
  e15685d8
04 Mar, 2026 13 commits
- fix: TRT-LLM multimodal preprocessor - revert to the old... · bb2ca1a9
  Michal Guzek authored Mar 04, 2026
```
fix: TRT-LLM multimodal preprocessor - revert to the old default_multimodal_input_loader for the embeddings case (#6840)
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
```
  bb2ca1a9
- chore: add mypy typing to frontend (#6861) · 2128d48a
  jh-nv authored Mar 04, 2026
  
  2128d48a
- chore: add mypy to planner and mocker (#6862) · 63e7b7da
  jh-nv authored Mar 04, 2026
  
  63e7b7da
- chore: add mypy typing for profiler (#6863) · 6216ae55
  jh-nv authored Mar 04, 2026
  
  6216ae55
- chore: add mypy for router (#6864) · a96f5d73
  jh-nv authored Mar 04, 2026
  
  a96f5d73
- fix: remove empty multimodal input to avoid invalid uuid check in vLLM (#6853) · 2ae9c290
  GuanLuo authored Mar 04, 2026
```
Signed-off-by: Guan Luo <41310872+GuanLuo@users.noreply.github.com>
```
  2ae9c290
- fix: strip None args when profiler generating configs (#6882) · 19e8a165
  Hongkuan Zhou authored Mar 04, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  19e8a165
- fix(trtllm): remove deprecated beam_width from health check payload (#6879) · bb79b7de
  Tzu-Ling Kan authored Mar 04, 2026
```
Signed-off-by: tzulingk <tzulingk@nvidia.com>
```
  bb79b7de
- feat: multi-image in request support for sglang backend (#6068) · adc95380
  Wang, Yi authored Mar 05, 2026
```
Signed-off-by: Wang, Yi <yi.a.wang@intel.com>
```
  adc95380
- docs: add readme in component directory (#6699) · 203249e1
  jh-nv authored Mar 04, 2026
  
  203249e1
- fix: refactor profiler's dgd generation workflow to correctly generate mocker config (#6848) · 7c1a6238
  Hongkuan Zhou authored Mar 04, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  7c1a6238
- fix(frontend): vllm processor works with stream_interval > 1 (#6816) · 4e1bd700
  Graham King authored Mar 04, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  4e1bd700
- chore: better error message for planner sweeping mode (#6844) · d688aa68
  Hongkuan Zhou authored Mar 03, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  d688aa68
03 Mar, 2026 6 commits
- fix: properly setup and register vLLM worker for external / hybrid load... · 90d74637
  GuanLuo authored Mar 03, 2026
```
fix: properly setup and register vLLM worker for external / hybrid load balancing. Update launch script (#6695)
Signed-off-by: Guan Luo <41310872+GuanLuo@users.noreply.github.com>
```
  90d74637
- chore(kv-router): deduplicate KvRouterConfig args into shared ArgGroup (#6805) · 9f3b7b33
  Yan Ru Pei authored Mar 03, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  9f3b7b33
- feat: use model_fields_set to distinguish ttft/itl default usage (#6814) · e4601811
  hhzhang16 authored Mar 03, 2026
```
Signed-off-by: Hannah Zhang <hannahz@nvidia.com>
```
  e4601811
- fix(frontend): Update vllm processor for vllm 0.16 (#6799) · 56212b49
  Graham King authored Mar 03, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  56212b49
- feat: FlexKV integration in Dynamo (#5858) · 4d7c9845
  Yue Yu authored Mar 03, 2026
```
Signed-off-by: zhuofan1123 <zhuofanl@nvidia.com>
Co-authored-by: zhuofan1123 <zhuofanl@nvidia.com>
```
  4d7c9845
- fix: Fix E + PD Multimodal Flow in trtllm (#6726) · 47ed1227
  Indrajit Bhosale authored Mar 02, 2026
```
Signed-off-by: Indrajit Bhosale <iamindrajitb@gmail.com>
```
  47ed1227