v0.22.0
Immutable
release. Only release title and notes can be modified.
What's Changed
- fix(memory): scope-aware recall isolation by @cniska in #274
- feat(context): trace window drops and add gap notice by @cniska in #275
- feat(trace): surface sink failures in the transcript by @cniska in #276
- fix(lifecycle): valid done completes, error shown once by @cniska in #277
- refactor(chat): unify TUI and run-mode output rendering by @cniska in #278
- feat(trace): count recall probes, show memory observations by @cniska in #279
- fix(trace): fix budget_exhausted_count field name by @cniska in #280
- fix: renderer solidify + tool-calling by @cniska in #281
- feat(chat): live status line by @cniska in #282
- test(tui): transcript-integrity VT harness by @cniska in #283
- fix: dogfood-surfaced chat and agent fixes by @cniska in #284
- fix(lifecycle): reject a done with no final answer by @cniska in #285
- fix(lifecycle): don't gate completion on tool errors by @cniska in #286
- refactor(tool-session): drop failure breaker by @cniska in #287
- fix(lifecycle): noop carries model text by @cniska in #288
- refactor(skills): replace keyword triggers with roster by @cniska in #289
- feat(agent): prose-first narration voice by @cniska in #290
- fix(lifecycle): user-facing completion errors by @cniska in #291
- fix(tui): truncate tool preview lines at width by @cniska in #292
- fix(agent): keep structure narration in prose by @cniska in #293
- test(tui): width-aware VT with deferred wrap by @cniska in #294
- test(tui): harden renderer tests by @cniska in #295
- fix(tui): nest chat diffs under sub-headers by @cniska in #296
- refactor: remove dead and unreachable code by @cniska in #297
- feat(skills): update built-in skills from agnostic set by @cniska in #298
- refactor(lifecycle): collapse duplicate completion gate by @cniska in #301
- feat(trace): count duplicate discovery calls by @cniska in #302
- fix(lifecycle): stop GPT signal-as-text leak by @cniska in #303
- refactor(memory): drop deprecated textEmbeddingModel by @cniska in #304
- fix(shell): scrub control sequences from output by @cniska in #305
- test(tool-utils): isolate git config in int tests by @cniska in #306
- fix(render): stop duplicating frozen scrollback by @cniska in #307
- chore: drop unused ast-grep grammar packages by @cniska in #308
- fix(shell): stop forcing color in subprocess env by @cniska in #309
- fix(chat): render stderr in same dim as stdout by @cniska in #310
- fix(agent): retain tool results across steps by @cniska in #311
- fix(agent): force tool choice for gateway GPT by @cniska in #312
- fix(agent): anchor the narration preamble rule by @cniska in #313
- feat(chat): show active skills by @cniska in #314
- fix(chat): transcript render and resume integrity by @cniska in #316
- fix(render): panel duplication on repeated resize by @cniska in #317
- fix(chat): keep blocked-turn reason on resume by @cniska in #318
- fix(chat): stop false prompt on blocked turns by @cniska in #319
- fix(chat): render blocked reason after prose by @cniska in #320
- fix(skill): seed session skills from request by @cniska in #321
- fix(skill): show active skill live during turn by @cniska in #322
- refactor(lifecycle): native end_turn completion by @cniska in #323
- refactor(agent): rework soul and completion handling by @cniska in #324
- refactor(web): one-line search preview by @cniska in #325
- feat(tui): render-layer line clipping by @cniska in #327
- refactor(tui): unify tool-output render path by @cniska in #326
- refactor(lifecycle): per-turn cap, drop reminders by @cniska in #328
- fix(chat): smooth bursty stream rendering by @cniska in #330
- refactor(code): drop multi-file code-edit by @cniska in #331
- feat(auth): OpenAI subscription auth by @cniska in #333
- feat(skill): model-owned skill deactivation by @cniska in #332
- feat(cli): global provider keys, vercel parity by @cniska in #334
- refactor(cli): unify provider authentication by @cniska in #335
- refactor(paths): unify global xdg layout by @cniska in #336
- fix(lifecycle): order tool output through fullStream by @cniska in #337
- fix(benchmark): refresh competitor metrics by @cniska in #338
Full Changelog: v0.21.0...v0.22.0