release: rust-v0.114.0-cometix
New Features
- Added an experimental code mode for more isolated coding workflows. (openai#13418)
- Added an experimental hooks engine with
SessionStartandStophook events. (openai#13276) - WebSocket app-server deployments now expose
GET /readyzandGET /healthzfor easier health checks. (openai#13782) - Added a config switch to disable bundled system skills entirely. (openai#13792)
- Handoffs now carry realtime transcript context for better continuity. (openai#14132)
- Improved the
$mention picker by clearly labeling Skills, Apps, and Plugins. (openai#14147, openai#14163)
Improvements
- Mark incomplete resumed turns interrupted when idle (openai#14125)
- Reuse McpToolOutput in McpHandler (openai#14229)
- Expose strongly-typed result for exec_command (openai#14183)
- Refactor tool output into trait implementations (openai#14152)
- tui: only show fast status for gpt-5.4 (openai#14135)
- Implemented thread-level atomic elicitation counter for stopwatch pausing (openai#12296)
Bug Fixes
- Fix Linux tmux segfault in user shell lookup (openai#13900)
- fix(core) default RejectConfig.request_permissions (openai#14165)
- fix: keep permissions profiles forward compatible (openai#14107)
- fix(protocol): preserve legacy workspace-write semantics (openai#13957)
- fix(core) RequestPermissions + ApplyPatch (openai#14055)
- [apps] Fix apps enablement condition (openai#14011)
Changelog
- openai#13418 Add code_mode experimental feature @da616136c
- openai#13276 start of hooks engine @244b2d53f
- openai#13782 add health endpoints for websocket server @66e71cce1
- openai#13792 support disabling bundled system skills @0c33af774
- openai#14132 Use realtime transcript for handoff context @2e24be213
- openai#14147 make dollar-mention clarify item category @a5af11211
- openai#14163 sort plugins first in $ menu @a3cd9f16f
- openai#14125 Mark incomplete resumed turns interrupted @b33edebd6
- openai#14229 Reuse McpToolOutput in McpHandler @46e6661d4
- openai#14183 Expose strongly-typed result for exec_command @e52afd28b
- openai#14169 Move exec command truncation into ExecCommandToolOutput @a9ae43621
- openai#14167 Export tools module into code mode runner @710682598
- openai#14165 fix(core) default RejectConfig.request_permissions @772259b01
- openai#14157 Enforce single tool output type @d71e04269
- openai#14152 Refactor tool output into trait implementations @aa04ea6bd
- openai#14135 tui: only show fast status for gpt-5.4 @63597d1b2
- openai#14118 RejectConfig for request_permissions @6da84efed
- openai#14107 keep permissions profiles forward compatible @1165a16e6
- openai#14055 RequestPermissions + ApplyPatch @c1defcc98
- openai#14136 dedicated types for responsesapi web search tool config @d309c102e
- openai#14009 Persist request_permission data across turns @d241dc598
- openai#14011 Fix apps enablement condition @566e4cee4
- openai#14000 Delay pending cleanup until task aborts @79307b793
- openai#13957 preserve legacy workspace-write semantics @b0cbc25a4
- openai#13900 Fix Linux tmux segfault @9a501ddb0
- openai#13603 Fix release-mode integration test compiler failure @b90921eba
- openai#12296 thread-level atomic elicitation counter @c6343e064