Commit Graph

Select branches

Hide Pull Requests

main

zhou

Mono Color

• b8e3293334
  test: skip test_sglang_indexers_sync in nightly only (DYN-2784) (#8539) main Tushar Sharma 2026-04-22 22:47:08 -0700
• 6b452a14c7
  chore(mocker): drop the Destroy variant (#8588) Yan Ru Pei 2026-04-22 21:44:29 -0700
• 241b254d83
  chore(profiler): reshape replay_optimize into DGDR-aligned Pydantic specs (#8585) Yan Ru Pei 2026-04-22 20:13:25 -0700
• d401c1a4c5
  refactor: Removes obsolete TRT-LLM request-abort disable wiring and re-enables cancellation test coverage (#8562) Jacky 2026-04-22 16:28:35 -0700
• 6ba534a63b
  fix(runtime): stop unconditional NATS connection with --discovery-backend file (#8398) Biswa Panda 2026-04-22 15:11:32 -0700
• 04ea38c310
  fix(parsers): back-port sglang kimi-k2 tool-call and reasoning detection (#8532) Nikita 2026-04-23 00:02:26 +0200
• 48911230c2
  test(fault_tolerance): replace hand-rolled SSE parser with openai SDK (#8536) Dmitry Tokarev 2026-04-22 17:10:01 -0400
• a2a2753d05
  fix: Remove double tokenization in EPP Integration fixes [DYN-2076] (#8093) atchernych 2026-04-22 14:09:43 -0700
• 4d5db80ae6
  chore: Upgrade ffmpeg to v8.1 (#8452) Graham King 2026-04-22 15:53:36 -0400
• 1aa5f219fe
  fix: remove the tool calling flaky test as the small model is not stable in generating Json format (#8523) Richard Huo 2026-04-22 12:29:54 -0700
• b441e26a6a
  feat: Interface for shared kv cache handling in kv routing (#7536) jthomson04 2026-04-22 12:10:32 -0700
• 1b0334c8ff
  feat: add TRTLLM text-to-image support (#8200) GuanLuo 2026-04-22 12:07:59 -0700
• 5135c321b0
  doc: update metrics doc regarding frontend staged gauges (#8459) Jie Hao 2026-04-22 14:51:18 -0400
• 69823d72eb
  feat(profiler): add replay objective knob (#8518) Yan Ru Pei 2026-04-22 11:38:29 -0700
• bce060d280
  fix(operator): drop Recreate override on restore-target Deployments (#8434) Schwinn Saereesitthipitak 2026-04-22 11:34:07 -0700
• 9321c2e5e9
  test: Explicit None token thresholds disable Router initiated Rejection (#8477) Jacky 2026-04-22 11:25:22 -0700
• 2a95ef631a
  fix(responses): align wire shape with OpenResponses spec + add compliance CI (#8283) ishandhanani 2026-04-22 13:15:39 -0500
• 4410a2c552
  ci: add nightly option to release pipeline (#8465) Anant Sharma 2026-04-22 14:10:41 -0400
• e79c4932c4
  feat: Add vLLM start/stop profile endpoints (#8068) leo-cf-tian 2026-04-22 13:57:00 -0400
• 81819fbc10
  fix(vllm): initialize failover flock in prefill worker (#8367) mohammedabdulwahhab 2026-04-22 10:56:27 -0700
• c4964f55c2
  feat(vllm): multinode elastic EP support via Ray cluster + concurrent scale fix (#8216) Tzu-Ling Kan 2026-04-22 12:54:18 -0500
• e041ccfcb1
  ci: remove unused SGLang NIXL ref (#8472) Alec 2026-04-22 10:16:00 -0700
• 36b4208e5e
  refactor(mocker): replace vllm block manager with kvbm-logical (#8451) Yongming Ding 2026-04-22 09:59:21 -0700
• 1dc0975b4b
  fix(runtime): reset canary health check timer on request arrival and per streaming chunk (#8467) Thomas Montfort 2026-04-22 09:54:34 -0700
• cf92be04bb
  docs: update docs for 1.1.0 DGDR (#8511) hhzhang16 2026-04-22 12:05:28 -0400
• 73a9a53f64
  feat(router): branch sharded kv indexer (#7859) Janelle Cai 2026-04-22 12:03:29 -0400
• af32579e89
  fix(vllm): cap gpu_memory_utilization for encoder-only vision model load (#8466) Qi Wang 2026-04-22 08:59:29 -0700
• 2cd4288a59
  fix(planner): match MDC component field against backend default, not DGD key (#8489) Hongkuan Zhou 2026-04-22 08:53:33 -0700
• dfd3aa99e4
  feat: address MR comments from TRTLLM/GMS integration (#8255) hhzhang16 2026-04-22 11:36:37 -0400
• d07288c542
  fix(operator): run GPU discovery when any hardware field is missing (#8267) Dr. Stefan Schimanski 2026-04-22 17:07:32 +0200
• bae41d44f2
  docs(agents): improve tool/reasoning parser docs (#8497) Keiven C 2026-04-22 08:06:20 -0700
• f53fa64c24
  fix: DIS-1768 emit structured SSE error + [DONE] on mid-stream fault (#8430) Keiven C 2026-04-22 08:02:10 -0700
• 9bf9709a65
  refactor(tests): consolidate diffusion payload classes into shared payloads.py (#8435) Krishnan Prashanth 2026-04-22 08:04:43 -0500
• d932d3b4ca
  fix(kvbm): Fix consolidator e2e tests (#8464) Kris Hung 2026-04-22 19:15:37 +0800
• 34acb1f443
  fix(sglang): populate total_kv_blocks via canonical Engine scheduler path (#8439) ishandhanani 2026-04-22 00:44:18 -0500
• ffa6e84afa
  feat: garbage collection tcp machine + read from shared place (#8420) Michael Feil 2026-04-21 22:07:57 -0700
• 73e0a2f2ef
  chore(mocker): fold workload concurrency cap into WorkloadDriver (#8446) Yan Ru Pei 2026-04-21 21:19:52 -0700
• 4833e29bdf
  feat(kv-router): warn on duplicate store replays (#8473) Yan Ru Pei 2026-04-21 18:40:45 -0700
• e9db304c0a
  fix(profiler): pass total_gpus to AIC enumerate so candidates respect budget (#8479) Hongkuan Zhou 2026-04-21 18:35:32 -0700
• d250fab987
  docs(observability): fix DYN_SYSTEM_PORT default in health-checks table (#8478) Jie Hao 2026-04-21 21:06:18 -0400
• 5c35805e46
  docs(profiler): clarify when planner-profile-data ConfigMap is emitted [DYN-2751] (#8486) Hongkuan Zhou 2026-04-21 17:31:03 -0700
• 6d44b904a5
  fix(planner): DYN-2776 multi-DGD + GlobalPlanner scaling and readiness (#8482) Hongkuan Zhou 2026-04-21 17:20:40 -0700
• 77a0051a6e
  fix: Delete SQL tool call test for now since small models could occasionally generate malformed arguments with guided decoding (#8474) Richard Huo 2026-04-21 16:46:27 -0700
• 8af634ba1b
  feat(global-router): add aggregated mode with TTFT x ITL pool selection (#8044) jthomson04 2026-04-21 16:11:10 -0700
• 7ffc699fe7
  fix(discovery): send ModelUpdate::Added after add_worker_set completes (#8242) MatejKosec 2026-04-21 15:40:48 -0700
• 1c9412d27c
  feat(lpu): add opaque engine_data field to nvext response with per-request opt-in via extra_fields Biswa Panda 2026-04-21 15:38:05 -0700
• 3f83c5974a
  fix(vllm): classify disagg decode-engine queued requests correctly in FPM (#8471) Hongkuan Zhou 2026-04-21 15:21:32 -0700
• d4ac0b59a8
  chore: rm kimi kvbm recipe | add qwen3 kvbm recipe (#8475) Ziqi Fan 2026-04-21 15:15:28 -0700
• 55a949cb66
  feat(profiler): wire mocker-rapid to direct AIC flags, drop profiler AIC interp (#8455) Hongkuan Zhou 2026-04-21 13:53:45 -0700
• cc583b2fe1
  test: stabilize nightly — skip engine-init failures, convert xfails to skips, fix http URL validation regression (#8443) Dmitry Tokarev 2026-04-21 16:10:48 -0400
• 9514236c87
  chore: upgrade modelexpress version (#8336) Nate Mailhot 2026-04-21 12:51:11 -0700
• ddd19a68ac
  test: skip test_router_decisions_sglang_multiple_workers (DYN-2784) (#8441) Tushar Sharma 2026-04-21 12:13:36 -0700
• eb6c65d5a8
  chore: copy attribution and license into planner image (#8453) Tushar Sharma 2026-04-21 12:07:07 -0700
• df888f6860
  docs: document pipeline time budgets for lifecycle markers (#8374) Tushar Sharma 2026-04-21 11:35:51 -0700
• b688597740
  feat(tests): add --models-dir flag for read-only HF cache (#8362) Ran Rubin 2026-04-21 21:24:00 +0300
• 4b3cd459df
  feat(kv-router): add start positions to flat KV events (#8426) Yan Ru Pei 2026-04-21 10:47:33 -0700
• e8ecf6ff5e
  fix(planner): use dynamo-planner image for profiler job and planner pods [DYN-2733][DYN-2746] (#8407) Hongkuan Zhou 2026-04-21 10:15:36 -0700
• f923777e9a
  fix: replace torch.load with safetensors and enable Rust frontend media decoding for TRT-LLM multimodal (#8295) Indrajit Bhosale 2026-04-21 09:51:36 -0700
• fd361c82f5
  Format fix (#8440) Dmitry Tokarev 2026-04-21 12:31:58 -0400
• 6f87f93227
  fix(mocker): cancel replay router scheduler tasks (#8429) Yan Ru Pei 2026-04-21 09:25:30 -0700
• e6ce4db3b2
  test(fault_tolerance): skip flaky sglang migration NATS combo, drop stale graceful_shutdown xfail (#8404) Dmitry Tokarev 2026-04-21 12:20:02 -0400
• d795e1a415
  test(serve): skip mm_agg_qwen2-audio-7b vLLM deployment on post-merge (#8411) Dmitry Tokarev 2026-04-21 12:15:35 -0400
• ec4ed33dcb
  ci: update runners (#8433) Ran Rubin 2026-04-21 18:40:32 +0300
• 8679faa35d
  test(sglang): add pre_merge diffusion CI smoke tests for T2I and T2V (#8409) Krishnan Prashanth 2026-04-21 04:53:11 -0700
• 19ecf46f0e
  docs(operator): clarify Interconnect and RDMA field semantics in HardwareSpec (#8300) Dr. Stefan Schimanski 2026-04-21 13:32:02 +0200
• 5103efdb64
  fix(vllm): include image geometry in multimodal hash preimage (#8341) Neelay Shah 2026-04-20 23:44:21 -0700
• ca15a79da7
  fix(snapshot): replace PID-1 SIGUSR1/SIGCONT contract with file sentinels (#8403) Schwinn Saereesitthipitak 2026-04-20 22:58:18 -0700
• 8fba4f568d
  fix: Validate multimodal media URLs and load paths (#8282) zhongdaor-nv 2026-04-20 22:19:47 -0700
• f701319e69
  chore(mocker): panic loudly on scheduler invariant violations (#8406) Yan Ru Pei 2026-04-20 20:38:54 -0700
• 781e410051
  docs(recipes): clarify Qwen3-235B DEEPGEMM requires Blackwell (SM100+) (#8410) Yuewei Na 2026-04-20 19:03:51 -0700
• 508aed8406
  fix(planner/tests): scope test_advisory_mode stubs so they don't leak (#8418) Dmitry Tokarev 2026-04-20 20:43:37 -0400
• d59b56be8b
  chore(deps): bump vLLM to 0.19.1 (#8370) Ayush Agarwal 2026-04-20 16:46:07 -0700
• 47cfbd46c9
  fix(parsers): recover Kimi K2 tool calls when section_end is missing (#8208) Keiven C 2026-04-20 15:28:13 -0700
• adfc02d58d
  feat: (Router) Guard threshold-based rejection logic when CLI thresholds are unset (#8333) Karen Chung 2026-04-20 15:21:17 -0700
• 8fd7de9a35
  fix: update the tool calling functionalities for sglang frontend processor to match with the latest sglang implementation (#8269) Richard Huo 2026-04-20 14:57:24 -0700
• 01688850eb
  fix(operator/deploy/tilt): build helm dependencies automatically (#8155) Dr. Stefan Schimanski 2026-04-20 23:21:20 +0200
• 303520f8cf
  fix(container): preserve file permissions in SGLang dev venv setup (#8408) Krishnan Prashanth 2026-04-20 14:17:37 -0700
• cdb40ec681
  fix(kv-router): allow unit block size in slot tracking (#8395) Yan Ru Pei 2026-04-20 13:29:34 -0700
• e71f1d2b44
  fix: llava EPD deployment. add llava coverage (#8330) GuanLuo 2026-04-20 13:17:01 -0700
• c68d159fe2
  fix(operator): reconcile DynamoGraphDeployment on PodCliqueScalingGroup status changes (#8328) Julien Mancuso 2026-04-20 14:04:14 -0600
• 4419009471
  fix(planner): backfill max_num_batched_tokens from discovery for VirtualConnector (#8042) jthomson04 2026-04-20 12:54:38 -0700
• 5c8f8ffbf7
  docs(mocker): clarify mocker image placeholder (#8274) Ryan McCormick 2026-04-20 12:49:52 -0700
• 846ce71a12
  fix(protocols): always serialize chat message content (null when absent) (#8372) Ryan McCormick 2026-04-20 12:21:44 -0700
• 7b979c0012
  fix: Add accelerate to sglang runtime image for diffusion workers (#8357) Krishnan Prashanth 2026-04-20 12:12:41 -0700
• 2e3d4a924a
  ci: expose cpu_parallel_mode and run planner tests with -n 0 (#8378) Dmitry Tokarev 2026-04-20 15:11:52 -0400
• 8291bfcc95
  docs(vllm-omni): note ARM64 not supported; flag GLM-Image as experimental (#8373) Ayush Agarwal 2026-04-20 12:09:54 -0700
• 073da7b1b1
  fix(planner): normalize model_name case in KubernetesConnector comparisons (#8384) Hongkuan Zhou 2026-04-20 11:58:09 -0700
• da7d6724d4
  fix(operator): emit Kubernetes syntax from LWS deployer so kubelet expands the leader hostname in direct-python container args (#8369) Julien Mancuso 2026-04-20 12:40:15 -0600
• 9054c882aa
  docs: add --prerelease=allow to sglang install commands (#8358) Krishnan Prashanth 2026-04-20 11:23:06 -0700
• 2618812218
  feat(planner): own AIC interpolation; fix MoE-DEP bugs in rapid mode (#8335) Hongkuan Zhou 2026-04-20 11:00:53 -0700
• 4ef2894067
  fix(responses): own input chain in dynamo-protocols to accept Codex/Agents-SDK shapes (#8275) ishandhanani 2026-04-20 12:30:53 -0500
• c388483aee
  feat(planner/replay): KV reuse awareness in load + throughput scaling (#8314) Hongkuan Zhou 2026-04-20 10:24:12 -0700
• db124db050
  ci: set skopeo runner to be prod-skopeo-v1 (#8347) Ran Rubin 2026-04-20 20:23:16 +0300
• f437c8cf44
  feat(frontend): gate nvext response metadata behind extra_fields (#8252) Biswa Panda 2026-04-20 09:47:06 -0700
• 0c98d9a152
  feat: add staged frontend gauges (#8162) Jie Hao 2026-04-20 12:21:21 -0400
• 7f58231bce
  docs: refresh quickstart to v1.0.1 and document partial v1.1.0-dev releases (#8364) Dan Gil 2026-04-20 10:44:40 -0500
• f460f4e25e
  feat: vLLM multinode elastic EP scaling support test (#8183) Tzu-Ling Kan 2026-04-20 09:49:21 -0500
• 1d154dcbec
  fix(sglang): add enable_trace to diffusion worker ServerArgs stub (#8332) Krishnan Prashanth 2026-04-20 03:02:50 -0700
• 714b423439
  fix: propagate vLLM --stream-interval to Dynamo frontend (#8101) Roger Wang 2026-04-19 23:44:27 -0700
• d94b350d2c
  chore(kv-router): make sequences stop doing token math (#8260) Yan Ru Pei 2026-04-17 15:20:34 -0700