Commits · bba26d55b76bef03b00e5edd7bc0dfe385720804 · OpenDAS / dynamo

13 Apr, 2026 1 commit
- feat(sglang): add ephemeral KV session routing (#7665) · 9498f016
  ishandhanani authored Apr 13, 2026
```
Signed-off-by: Ishan Dhanani <ishandhanani@gmail.com>
```
  9498f016
09 Apr, 2026 1 commit
- perf: avoid cloning content parts in gather_multi_modal_data (#8009) · f49df565
  KrishnanPrash authored Apr 08, 2026
  
  f49df565
06 Apr, 2026 1 commit
- perf(multimodal): strip inline base64 from TCP transport in frontend decoding (#7895) · 6b8545fc
  Qi Wang authored Apr 06, 2026
```
Co-authored-by: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
```
  6b8545fc
02 Apr, 2026 3 commits
- feat: enable DP for GAIE (#7741) · 2b906504
  atchernych authored Apr 02, 2026
```
Signed-off-by: Anna Tchernych <atchernych@nvidia.com>
```
  2b906504
- fix: tool_choice=required bypasses format-specific parsers (#6821) (#7589) · 07af8dca
  Asad Shahid authored Apr 02, 2026
  
  07af8dca
- refactor(protocols): deprecate cache control (#7790) · c09ac697
  ishandhanani authored Apr 01, 2026
  
  c09ac697
01 Apr, 2026 1 commit
- refactor(2/3): rename dynamo-async-openai to dynamo-protocols (#7565) · b6a3b0c6
  ishandhanani authored Apr 01, 2026
  
  b6a3b0c6
31 Mar, 2026 1 commit

fix: strip null bytes before tokenization to prevent false positive prompt... · 023a299c

Albert Cheng authored Mar 31, 2026

fix: strip null bytes before tokenization to prevent false positive prompt rejection due to Rust/Python tokenizer divergence (#7694)
Signed-off-by: Albert Cheng (Engrg-Hardware 1) <albecheng@login-lyris01.lyris.clusters.nvidia.com>
Co-authored-by: Albert Cheng (Engrg-Hardware 1) <albecheng@login-lyris01.lyris.clusters.nvidia.com>

023a299c

30 Mar, 2026 1 commit
- refactor(1/3): move `nvext` to `dynamo-llm` and move `anthropic` to `dynamo-async-openai` (#7564) · 2887cd1c
  ishandhanani authored Mar 30, 2026
  
  2887cd1c
23 Mar, 2026 1 commit
- feat(perf): add pipeline instrumentation and feature gated NVTX markers (#6746) · 82794761
  Biswa Panda authored Mar 23, 2026
  
  82794761
20 Mar, 2026 2 commits
- feat(mocker): add offline trace replay mode [DYN-2502] (#7543) · c3908a36
  Yan Ru Pei authored Mar 20, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  c3908a36
- feat(llm): unify user-facing priority hints (#7492) · ebd06a1f
  Yan Ru Pei authored Mar 20, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  ebd06a1f
12 Mar, 2026 1 commit
- fix: respect per-request chat_template_kwargs for DeepSeek V3.2 thinking mode (#7286) · 745d7e4c
  brluo authored Mar 12, 2026
```
Co-authored-by: kangclzjc <kangz@nvidia.com>
```
  745d7e4c
11 Mar, 2026 2 commits
- fix: add force_nonempty_content for nemotron (#7225) · c95bfc2e
  Neal Vaidya authored Mar 11, 2026
```
Signed-off-by: Neal Vaidya <nealv@nvidia.com>
```
  c95bfc2e
- feat(anthropic): add thinking block support and preamble stripping to /v1/messages (#7137) · 012236ee
  MatejKosec authored Mar 10, 2026
```
Signed-off-by: Matej Kosec <mkosec@nvidia.com>
```
  012236ee
06 Mar, 2026 2 commits

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: 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 1 commit
- fix(llm): preserve reasoning content when tool-call starts mid-chunk (#6902) · 57e6a79f
  Graham King authored Mar 05, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  57e6a79f
03 Mar, 2026 1 commit
- feat: upgrade rust to v1.93.0 (#6802) · 57bdbdf1
  Ryan Olson authored Mar 03, 2026
```
Signed-off-by: Ryan Olson <rolson@nvidia.com>
```
  57bdbdf1
02 Mar, 2026 1 commit

feat: Full Anthropic Messages API cache_control support (top-level, per-block,... · 4d3e1ae3

MatejKosec authored Mar 02, 2026


feat: Full Anthropic Messages API cache_control support (top-level, per-block, system block arrays) (#6629)
Signed-off-by: Matej Kosec <mkosec@nvidia.com>

4d3e1ae3

27 Feb, 2026 1 commit
- feat: wire nvext.cache_control TTL-based pinning through Dynamo router (#6213) · f1d5c95a
  ishandhanani authored Feb 26, 2026
```
Co-authored-by: Claude <noreply@anthropic.com>
```
  f1d5c95a
26 Feb, 2026 1 commit
- fix: use force_reasoning parser for nemotron_nano (#6288) · a1766e4a
  Neal Vaidya authored Feb 25, 2026
```
Signed-off-by: Neal Vaidya <nealv@nvidia.com>
```
  a1766e4a
25 Feb, 2026 3 commits
- feat: Tiktoken support (#6460) · 21fce9ba
  Nikita authored Feb 25, 2026
```
Signed-off-by: Nikita Sukharev <kaonael@gmail.com>
```
  21fce9ba
- chore: plumb allowed_worker_ids through RoutingHints (#6580) · ebaf048d
  Yan Ru Pei authored Feb 25, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
```
  ebaf048d
- feat: Kimi K2/K2.5 tool and reasoning parsers (#6407) · d1bd210f
  Nikita authored Feb 25, 2026
```
Signed-off-by: Nikita Sukharev <kaonael@gmail.com>
```
  d1bd210f
24 Feb, 2026 1 commit
- feat: Standardized Dynamo Error Type (#6303) · c8276cd2
  Jacky authored Feb 23, 2026
```
Signed-off-by: Jacky <18255193+kthui@users.noreply.github.com>
```
  c8276cd2
18 Feb, 2026 1 commit
- feat: plumb request priority through sglang and vllm handlers (#6348) · ee0218b5
  Yan Ru Pei authored Feb 17, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: Cursor <cursoragent@cursor.com>
```
  ee0218b5
15 Feb, 2026 1 commit

feat: speculative prefill (#6230) · 4ede59a2

Yan Ru Pei authored Feb 15, 2026


Signed-off-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: Janelle Cai <jcai18@mit.edu>

4ede59a2

13 Feb, 2026 2 commits
- feat: Decomposed pipeline for EPP integration [DEP-730] (#5446) · 52271760
  atchernych authored Feb 12, 2026
```
Signed-off-by: Anna Tchernych <atchernych@nvidia.com>
```
  52271760
- feat: router priority queue (#6010) · bc514fbe
  Yan Ru Pei authored Feb 12, 2026
```
Signed-off-by: PeaBrane <yanrpei@gmail.com>
Co-authored-by: ishandhanani <82981111+ishandhanani@users.noreply.github.com>
```
  bc514fbe
11 Feb, 2026 1 commit
- feat: Metric for detokenization latency (#6160) · adaf1a39
  Graham King authored Feb 11, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  adaf1a39
10 Feb, 2026 3 commits
- chore: remove media-nixl feature (#5940) · bf9e6d04
  milesial authored Feb 10, 2026
```
Signed-off-by: Alexandre Milesi <milesial@users.noreply.github.com>
```
  bf9e6d04
- feat(frontend): Use vllm for pre and post processing (#5544) · 4f99451b
  Graham King authored Feb 10, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  4f99451b
- feat: Add metric tokenizer_latency_ms (#6092) · f1bcb175
  Graham King authored Feb 10, 2026
```
Signed-off-by: Graham King <grahamk@nvidia.com>
```
  f1bcb175
09 Feb, 2026 1 commit
- feat: text to image vLLM Omni (#5912) · 9f76d060
  Ayush Agarwal authored Feb 09, 2026
```
Signed-off-by: ayushag <ayushag@nvidia.com>
```
  9f76d060
05 Feb, 2026 1 commit
- feat(lora): implement HRW-based LoRA allocation system (#5992) · 953e5d7b
  Biswa Panda authored Feb 05, 2026
  
  953e5d7b
04 Feb, 2026 2 commits
- feat: Add per-worker Prometheus metrics for router load monitoring (#5842) · 0c0336e6
  Hongkuan Zhou authored Feb 04, 2026
```
Signed-off-by: hongkuanz <hongkuanz@nvidia.com>
```
  0c0336e6
- fix: fuse unfold streams to prevent panics from poll after termination (#5872) · 7beb5f1a
  Janelle Cai authored Feb 04, 2026
  
  7beb5f1a
03 Feb, 2026 1 commit
- feat(lora): Add lora_name tracking to scheduling and sequence management (#5875) · 373e76c1
  Biswa Panda authored Feb 02, 2026
  
  373e76c1
21 Jan, 2026 1 commit
- feat: remove ser/deser in gather_multi_model_data (#5485) · 0f137de4
  milesial authored Jan 21, 2026
```
Signed-off-by: Alexandre Milesi <milesial@users.noreply.github.com>
```
  0f137de4