All pull requests

Pull requests

• Validate predictive structured JSON bounds

  #1325·Akshay Sonawane (apsonawane) opened Nov 11, 2026Contributor·

  ·
  6

• Bump oatpp to 1.3.1 to fix Expect: 100-continue deadlock

  #1324·bmehta001 opened Nov 11, 2026Contributor·

  ·
  2

• Remove stale GenAI LSan suppressions and fix tool-call parsing UAF

  #1323·bmehta001 opened Nov 11, 2026Contributor·

  ·
  2

• Target .NET 10 in JS dotnet-mode restore

  #1321·bmehta001 opened Nov 10, 2026Contributor·

  ·
  2

• Honor per-model external prefill chunking capability

  #1320·Sayan Shaw (sayanshaw24) opened Nov 10, 2026Contributor·

  ·
  2

• Honor model-configured sampling defaults

  #1315·bmehta001 opened Nov 10, 2026Contributor·

  ·
  2

  

• Preserve images across conversation turns

  #1313·bmehta001 opened Nov 10, 2026Contributor·

  ·
  2

• [DO NOT MERGE] Avoid default external prefill chunking for Nemotron Bolt

  #1312·bmehta001 opened Nov 10, 2026Contributor·

  ·
  2

• implements bring your own model catalog feature via azureml registries

  #1311·Prathik Rao (prathikr) opened Nov 10, 2026Collaborator·

  ·
  35

• [Speech] Populate timestamped ASR results from GenAI postprocessing

  #1310·Mason Corey (themason2011) opened Nov 9, 2026Contributor·

  ·
  1

• Test retained chat history after edits and reordering

  #1309·Baiju Meswani (baijumeswani) opened Nov 9, 2026Collaborator·

  ·
  2

• No longer override generic GPU models to CUDA

  #1306·bmehta001 opened Nov 7, 2026Contributor·

  ·
  1

  2

  

• Add custom model HTTP transport foundation

  #1301·Ayman Bagabas (aymanbagabas) opened Nov 6, 2026Member·

  ·
  3

• Support native tool continuations

  #1296·Sayan Shaw (sayanshaw24) opened Nov 5, 2026Contributor·

  ·
  77

• Add GPU total-VRAM compatibility filtering to the native model catalog

  #1295·Selena Yang (selenayang888) opened Nov 5, 2026Contributor·

  ·
  12

• SDK v2: expose HTTP activity snapshots and atomic idle stop

  #1290·bmehta001 opened Nov 4, 2026Contributor·

  ·
  3

• Preserve tool-call validation errors when guided recovery fails

  #1282·bmehta001 opened Nov 4, 2026Contributor·

  ·
  1

  2

  

• Add on-device text-to-speech (text-to-speech) on the ORT GenAI VibeVoice model type

  #1271·Aymen (aymenfurter) opened Nov 1, 2026·

  ·
  14

• Bump the npm_and_yarn group across 2 directories with 3 updates

  dependenciesjavascript
  #1269·dependabot[bot] opened Oct 31, 2026Bot·

  ·
  8

• Add versioned cross-stack tool-calling qualification

  #1246·Sayan Shaw (sayanshaw24) opened Oct 26, 2026Contributor·

  ·
  16

• Expose SDK AudioModes for explicit audio workflow selection

  #1235·Mason Corey (themason2011) opened Oct 23, 2026Contributor·

  ·
  4

• Delegate detailed hardware and resource telemetry to GenAI

  #1229·bmehta001 opened Oct 22, 2026Contributor·

  ·
  2

• Whisper segment timestamps

  #1148·Joel Stidley (joelst) opened Sep 25, 2026·

  ·
  24

• Reclaim cached GenAI CUDA resources after the last CUDA model unloads

  #1110·aciddelgado opened Sep 15, 2026·

  ·
  9

• Add model package support for sdkv2

  #981·Xiaoyu Z (xiaoyu-work) opened Aug 11, 2026Member·

  ·
  29