Pull requests

• [Bug] Map vtc-basic users onto pods by a stable order, not list position

  area/gateway
  #3245·AarnavNoble opened Oct 21, 2026Contributor·

  ·
  1

  2

• [Bug] Keep pods of roles without pod groups when a PodAutoscaler targets a whole StormService

  area/orchestration
  #3244·hmdsefi opened Oct 21, 2026Contributor·

  ·
  1

  2

• [Bug] Republish a KV cache pod's address when it changes under the same name

  #3242·hmdsefi opened Oct 21, 2026Contributor·

  ·
  1

  2

• [Bug] Keep all of a KV cache pod's slot ranges when it owns the first and last slot

  #3240·hmdsefi opened Oct 21, 2026Contributor·

  ·
  1

  2

• [Gateway] Support PD routing for /v1/responses on vLLM and SGLang

  area/gatewayarea/websitekind/documentationtriage/needs-information
  #3234·Yang1032 opened Oct 21, 2026Contributor·

  ·
  2

• [Bug] Cap Playground max tokens for SGLang ModelClaims

  area/website
  #3233·bolubo opened Oct 21, 2026Contributor·

  ·
  2

• [Bug] Charge vtc-basic user tokens when the route goes through the blend

  area/gateway
  #3231·AarnavNoble opened Oct 20, 2026Contributor·

  ·
  1

  10

• [Bug] Let the auto-blend weight env vars be set to 0

  area/gatewaykind/documentation
  #3230·AarnavNoble opened Oct 20, 2026Contributor·

  ·
  1

  7

• [Bug] Check a node with the API server before a Once ModelWarmup fails it

  area/orchestration
  #3229·Zeyu-ZEYU opened Oct 20, 2026Contributor·

  ·
  9

• [Bug] Skip a prefill pod far above the pool mean at any pool size

  area/gatewayarea/websitekind/documentation
  #3224·AarnavNoble opened Oct 20, 2026Contributor·

  ·
  1

  6

• [Feat] Add bounded MEMORY backend for variable-length KV records

  area/runtime
  #3221·mynotwo opened Oct 20, 2026·

  ·
  8

• [Bug] Return an error status when the Redis cluster meta key is missing

  area/runtime
  #3218·wguo7 opened Oct 20, 2026·

  ·
  1

  3

• [Docs] Add MUSA S5000 inference samples

  area/installationkind/documentation
  #3216·googs1025 opened Oct 20, 2026Collaborator·4 of 5 tasks·

  ·
  3

• [Bug] Preserve input usage across Messages stream events

  area/gateway
  #3214·LingYi-Liang opened Oct 20, 2026Contributor·

  ·
  8

• [Bug] Enforce the default TPM for streamed requests

  area/gateway
  #3213·bolubo opened Oct 20, 2026Contributor·

  ·
  6

• [Bug] Apply updated GPU profiles to deployments registered by namespace

  area/runtime
  #3211·wguo7 opened Oct 20, 2026·

  ·
  1

  3

• [Bug] Decode prefix and query tokens from the packed token area

  area/runtime
  #3209·wguo7 opened Oct 20, 2026·

  ·
  1

  4

• [Bug] Keep ghost entry sizes so the S3FIFO ghost fifo is trimmed

  area/runtime
  #3207·wguo7 opened Oct 20, 2026·

  ·
  1

  3

• [Bug] Apply the job's client request timeout to batch inference requests

  area/batcharea/runtime
  #3205·wguo7 opened Oct 20, 2026·

  ·
  1

  1

• [Bug] Recover KV event replay across every remaining sequence gap

  #3203·Jackmeson1 opened Oct 20, 2026Contributor·

  ·
  1

  2

• [Bug] Run the E+PD encode stage when a replica inflight cap is set

  area/gateway
  #3200·theminjunchoi opened Oct 20, 2026Contributor·

  ·
  1

  1

• [Feat] Pass encoders' ec_transfer_params to the request that consumes them

  area/gatewayarea/installationarea/kv-cachearea/websitekind/documentation
  #3197·theminjunchoi opened Oct 20, 2026Contributor·

  ·
  3

• [Bug] Correct batch engine argument mappings for SGLang and vLLM

  area/batcharea/runtimekind/documentationtriage/needs-information
  #3191·GeorgiaPreda opened Oct 19, 2026·

  ·
  5

• [Bug] Prevent recursive TOS multipart uploads

  area/runtime
  #3184·ssfdgsg opened Oct 19, 2026·

  ·
  1

  4

• [Bug] Honor AIBRIX_PREFIX_CACHE_REMOTE_TOKENIZER_ENDPOINT and spread tokenize endpoint selection

  area/gatewaykind/documentation
  #3181·chiiydd opened Oct 19, 2026·

  ·
  1

  3