Commits · 651ef5b5063a70e1e80b4c111d3168370ff83aec · OpenDAS / dynamo

25 Feb, 2026 4 commits
- feat: throughput-metrics-source for SLA planner + GlobalPlanner disagg scaling (#6500) · 651ef5b5
  daiyaanarfeen authored Feb 25, 2026
```
Signed-off-by: Daiyaan <darfeen@nvidia.com>
Co-authored-by: Claude Sonnet 4.6 <noreply@anthropic.com>
```
  651ef5b5
- perf: Keep embeddings on GPU Embedding Sender in EPD pipeline + minor fixes (#6535) · 80955ef4
  Kris Hung authored Feb 25, 2026
  
  80955ef4
- feat(mocker): add KV cache transfer latency for disagg serving (#6504) · 5d958623
  Yongming Ding authored Feb 24, 2026
```
Signed-off-by: Yongming Ding <yongmingd@nvidia.com>
```
  5d958623
- feat: add ZMQ KV event publishing to mocker [DYN-2221] (#6528) · 65dc451d
  Yan Ru Pei authored Feb 24, 2026
```
Signed-off-by: Ru Pei <rupei@nvidia.com>
Signed-off-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  65dc451d
24 Feb, 2026 12 commits
- fix: restore E/P/D multimodal disagg serving and add Qwen3-VL-30B-A3B support (#6533) · 8e236375
  KrishnanPrash authored Feb 24, 2026
```
Signed-off-by: Krishnan Prashanth <kprashanth@nvidia.com>
Signed-off-by: KrishnanPrash <140860868+KrishnanPrash@users.noreply.github.com>
```
  8e236375
- feat: sglang to 0.5.9 + updated docs (#6518) · 6642e23e
  ishandhanani authored Feb 24, 2026
```
Co-authored-by: baihuitian <baihuitian.bht@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
```
  6642e23e
- chore: install vllm-omni in vllm container (#6458) · e94f4647
  Ayush Agarwal authored Feb 24, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
  e94f4647
- fix: frontend crash when using TRTLLM runtime image (#6481) · 35128b14
  Raghav Potluri authored Feb 24, 2026
```
Signed-off-by: Raghav Potluri <raghav.potluri21@gmail.com>
```
  35128b14
- chore: split JetStream subscriber into dedicated module and deprecate... · bc00ef38
  Yan Ru Pei authored Feb 24, 2026
```
chore: split JetStream subscriber into dedicated module and deprecate durable_kv_events [DYN-2203] (#6477)
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  bc00ef38
- feat: make KV cache events and routing LoRA-aware (#6517) · 75fea787
  Biswa Panda authored Feb 24, 2026
  
  75fea787
- feat: remove --connector flag for vLLM backend (LLM-90) (#6450) · 6d3b92f0
  Alec authored Feb 24, 2026
```
Signed-off-by: alec-flowers <aflowers@nvidia.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  6d3b92f0
- chore: replace --enforce-disagg with --decode-fallback, default to enforcing disagg (#6515) · c9ff6235
  Yan Ru Pei authored Feb 24, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  c9ff6235
- feat: add Planner schema to DGDR and Profiler input (#6463) · 659122ed
  hhzhang16 authored Feb 24, 2026
```
Signed-off-by: Hannah Zhang <hannahz@nvidia.com>
Co-authored-by: Sertaç Özercan <852750+sozercan@users.noreply.github.com>
```
  659122ed
- feat: embedding cache in agg vLLM node (#6153) · 33b3cb8a
  Qi Wang authored Feb 23, 2026
  
  33b3cb8a
- chore: Expose new kv_cache metrics from trtllm backend (#6469) · eb76a8b5
  Indrajit Bhosale authored Feb 23, 2026
  
  eb76a8b5
- feat: add --disaggregation-mode enum to vLLM backend (#6483) · 7893f268
  Alec authored Feb 23, 2026
```
Signed-off-by: alec-flowers <aflowers@nvidia.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  7893f268
23 Feb, 2026 6 commits
- feat: Remove Component from public (#6403) · 80cac7c1
  Tzu-Ling Kan authored Feb 23, 2026
```
Signed-off-by: tzulingk@nvidia.com <tzulingk@nvidia.com>
```
  80cac7c1
- feat(runtime): add hierarchical Model/WorkerSet architecture for multi-namespace support (#6054) · cb55766c
  Thomas Montfort authored Feb 23, 2026
```
Signed-off-by: tmontfort <tmontfort@nvidia.com>
```
  cb55766c
- chore: use config.modality directly for multimodal init and remove incorrect fallback value (#6471) · c40f3525
  jh-nv authored Feb 23, 2026
  
  c40f3525
- fix: prefix omni/diffusion CLI flags with --omni- to avoid vLLM collisions (#6476) · 43ce8608
  Alec authored Feb 23, 2026
```
Signed-off-by: alec-flowers <aflowers@nvidia.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  43ce8608
- fix(chrek): fix CI errors and replace SIGUSR2 with SIGKILL on checkpoint failure (#6478) · 673822ea
  Schwinn Saereesitthipitak authored Feb 23, 2026
  
  673822ea
- feat: Backend accept new requests during shutdown grace period (#6093) · ea86df29
  jh-nv authored Feb 22, 2026
```
Signed-off-by: Jacky <18255193+kthui@users.noreply.github.com>
Co-authored-by: Jacky <18255193+kthui@users.noreply.github.com>
```
  ea86df29
22 Feb, 2026 2 commits

feat: add --headless mode for multi-node TP/PP in dynamo.vllm (#6204) · 4ebb244b

Alec authored Feb 22, 2026


Signed-off-by: alec-flowers <aflowers@nvidia.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

4ebb244b

fix: add pre_merge markers to vLLM unit tests (#6482) · a9568455

Alec authored Feb 22, 2026


Signed-off-by: alec-flowers <aflowers@nvidia.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>

a9568455

21 Feb, 2026 2 commits
- fix: Multimodal flag was ignored for TRTLLM (#6468) · 8d30cd45
  William Zhang authored Feb 21, 2026
```
* Why?

Commit `5a67b246` refactored configs for the TRTLLM backend, breaking
`--modality multimodal`.

* What?

This commit fixes this bug, and adds a unit test verified to fail
without it.
```
  8d30cd45
- feat(chrek): external restore, signal-based IPC, and package refactor (#6286) · bb8fc8a4
  Schwinn Saereesitthipitak authored Feb 20, 2026
```
Co-authored-by: Dan Feigin <dfeigin@nvidia.com>
```
  bb8fc8a4
20 Feb, 2026 10 commits
- feat(frontend): Reduce Python-side overhead in the vLLM chat path (#6437) · f91b42b9
  Graham King authored Feb 20, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  f91b42b9
- feat: Configurable Request Cancellation abort passage to TRT-LLM (#6445) · 41d7d549
  Jacky authored Feb 20, 2026
```
Signed-off-by: Jacky <18255193+kthui@users.noreply.github.com>
```
  41d7d549
- fix: vllm omni image perf fix (#6451) · 7409bd3a
  Ayush Agarwal authored Feb 20, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
  7409bd3a
- feat: add embedding cache to pd worker (#6061) · c82fe888
  Qi Wang authored Feb 20, 2026
  
  c82fe888
- feat: Add v1beta1 DGDR API with conversion framework (#6352) · ebc61637
  hhzhang16 authored Feb 19, 2026
```
Signed-off-by: Jont828 <jt572@cornell.edu>
Signed-off-by: Hongkuan Zhou <hongkuanz@nvidia.com>
Signed-off-by: Hannah Zhang <hannahz@nvidia.com>
Co-authored-by: Jont828 <jt572@cornell.edu>
```
  ebc61637
- feat: default kv-events-config to empty (align with vLLM defaults) (#6404) · 7bbacce1
  Alec authored Feb 19, 2026
```
Signed-off-by: alec-flowers <aflowers@nvidia.com>
Co-authored-by: Claude Opus 4.6 <noreply@anthropic.com>
```
  7bbacce1
- chore: unify filesystem usage across all frameworks and workers (#6391) · d6c49779
  Ayush Agarwal authored Feb 19, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
  d6c49779
- feat: vllm omni text to video generation pipeline (#6104) · 6dd3ce2e
  Ayush Agarwal authored Feb 19, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
  6dd3ce2e
- feat: e2e mm aware kv cache routing support for trtllm backend (#5480) · 23de4e86
  zhongdaor-nv authored Feb 19, 2026
```
Signed-off-by: zhongdaor <zhongdaor@nvidia.com>
Signed-off-by: zhongdaor-nv <zhongdaor@nvidia.com>
Signed-off-by: Zhongdao Ren <zhongdaor@zhongdaor-mlt.client.nvidia.com>
Co-authored-by: Claude Sonnet 4.5 <noreply@anthropic.com>
Co-authored-by: Zhongdao Ren <zhongdaor@zhongdaor-mlt.client.nvidia.com>
```
  23de4e86
- feat: TRT-LLM multimodal preprocessor with backend media decoding (#5910) · 8a098a66
  Michal Guzek authored Feb 19, 2026
```
Signed-off-by: Michal Guzek <mguzek@nvidia.com>
```
  8a098a66
19 Feb, 2026 4 commits
- feat: resolve lora request for multimodal workers (#6399) · 026f361d
  Biswa Panda authored Feb 19, 2026
  
  026f361d
- feat: Add runtime.endpoint() method to eliminate namespace chaining (#6386) · 0ce3461a
  Tzu-Ling Kan authored Feb 19, 2026
```
Signed-off-by: tzulingk@nvidia.com <tzulingk@nvidia.com>
```
  0ce3461a
- feat: add model field to multimodal protocol for LoRA identity (#6382) · 4ba48b79
  Biswa Panda authored Feb 19, 2026
  
  4ba48b79
- feat: use embedding transfer classes for EPD (#6223) · a2a6917f
  GuanLuo authored Feb 19, 2026
```
Signed-off-by: Guan Luo <41310872+GuanLuo@users.noreply.github.com>
```
  a2a6917f