All pull requests

Pull requests

• fix(libsy): preserve retained sessions on cache updates

  #1190·Yushu-Li opened Oct 31, 2026·

  ·
  1

• feat(runner): accept subagents on an llm_classifier route

  #1182·chethanuk opened Oct 30, 2026Contributor·

  ·
  1

  3

• feat(routing): add optional N-target score pools to fallthrough

  #1172·sabhatinas opened Oct 29, 2026Contributor·

  ·
  4

  

• feat(openshell): add policy-driven routing middleware example

  #1169·afourniernv opened Oct 28, 2026Contributor·

  ·

• feat(translation): support Anthropic MCP calls and results

  #1162·dagardner-nv opened Oct 27, 2026Contributor·

  ·
  6

• feat(desktop): continue a session in another harness

  #1148·eric-liu-nvidia opened Oct 26, 2026Contributor·8/8·

  ·

• feat(go): report the compiled routing backend

  #1147·afourniernv opened Oct 26, 2026Contributor·

  ·
  3

• feat(desktop): record session launches and open harnesses in existing worktrees

  #1146·eric-liu-nvidia opened Oct 26, 2026Contributor·

  ·
  5

• feat(routing): speculate bounded tool calls with decision models

  #1144·nachiketb-nvidia opened Oct 26, 2026Contributor·

  ·
  1

• fix(llm-client): fall back on retryable pre-answer stream errors

  #1133·cpakkamisaac-sae opened Oct 23, 2026Contributor·

  ·
  1

  3

• ci(go): validate WASM routing bindings

  #1131·afourniernv opened Oct 23, 2026Contributor·

  ·

• feat(translation): support Responses MCP definitions

  #1125·dagardner-nv opened Oct 23, 2026Contributor·

  ·
  2

• feat(desktop): read harness transcripts for session hand-offs

  #1079·eric-liu-nvidia opened Oct 22, 2026Contributor·

  ·
  16

• fix(translation): keep Codex namespaces off calls to top-level tools

  #1075·v0ropaev opened Oct 22, 2026·

  ·
  4

• fix(translation): map OpenAI none and minimal reasoning effort for Anthropic

  #1074·v0ropaev opened Oct 22, 2026·

  ·
  4

• feat(libsy): add best-of-N action sampling

  #1068·bhuvan002 opened Oct 21, 2026Contributor·

  ·
  2

• feat(classifier): add cited requirement feedback to a fixed target

  #1064·sabhatinas opened Oct 20, 2026Contributor·

  ·
  1

• feat(translation): support Anthropic MCP definitions

  #1061·dagardner-nv opened Oct 20, 2026Contributor·

  ·
  5

• feat(llm-client): redact MCP credentials from upstream errors

  #1037·dagardner-nv opened Oct 19, 2026Contributor·

  ·
  8

• feat(classifier): add optional hints to a fixed target

  #1029·sabhatinas opened Oct 16, 2026Contributor·

  ·
  2

  

• feat(protocol): support MCP streaming content

  #1025·dagardner-nv opened Oct 16, 2026Contributor·

  ·
  3

• feat(routing): make composite classifier optional

  #1005·ryan-lempka opened Oct 15, 2026Collaborator·

  ·
  2

• feat(llm-client): send x-litellm-session-id for LiteLLM prompt caching

  #1001·eugenn opened Oct 15, 2026Contributor·

  ·
  11

• fix(translation): preserve Anthropic server tools during reconstruction

  #1000·Atharva-Kanherkar opened Oct 15, 2026Contributor·

  ·
  1

  3

• feat(runner): let targets share one model with different reasoning effort

  #991·sabhatinas opened Oct 14, 2026Contributor·

  ·
  18