# Changelog ## [Unreleased] ## 0.6.6 - 2026-11-02 - Adopt the strict Oxlint/Oxfmt engineering standard (v5.1.2) (#46) ([af7fa6d](https://github.com/fitchmultz/pi-tool-duration/commit/af7fa6de4f73c13c972f59e9749d775c05ee5e9a)) ### What's Changed - Adopt the strict Oxlint/Oxfmt engineering standard (v5.1.2) by @fitchmultz in https://github.com/fitchmultz/pi-tool-duration/pull/46 **Full Changelog**: https://github.com/fitchmultz/pi-tool-duration/compare/v0.6.5...v0.6.6 ## 0.6.5 - 2026-11-02 - chore(deps-dev): bump brace-expansion from 5.0.9 to 5.0.12 (#37) ([0710efb](https://github.com/fitchmultz/pi-tool-duration/commit/0710efb611c04032e7fe8ff0ddceac4360f8b28d)) ### What's Changed - chore(deps-dev): bump brace-expansion from 5.0.9 to 5.0.12 by @dependabot[bot] in https://github.com/fitchmultz/pi-tool-duration/pull/37 ### New Contributors - @dependabot[bot] made their first contribution in https://github.com/fitchmultz/pi-tool-duration/pull/37 **Full Changelog**: https://github.com/fitchmultz/pi-tool-duration/compare/v0.6.4...v0.6.5 ## 0.6.4 - 2026-11-01 - Use direct Flash for automatic release remediation (#42) ([eb6fbf5](https://github.com/fitchmultz/pi-tool-duration/commit/eb6fbf5765d1602512f622f123c8aaa76ec31b33)) ### What's Changed - Use direct Flash for necessary release remediation by @fitchmultz in https://github.com/fitchmultz/pi-tool-duration/pull/42 **Full Changelog**: https://github.com/fitchmultz/pi-tool-duration/compare/v0.6.3...v0.6.4 ## 0.6.3 - 2026-11-01 - Pin reviewed release source binding (#40) ([e34f502](https://github.com/fitchmultz/pi-tool-duration/commit/e34f50225957183c0cc9a2312d68076d4caa7717)) ### What's Changed - Bind automatic releases to the reviewed version by @fitchmultz in https://github.com/fitchmultz/pi-tool-duration/pull/40 **Full Changelog**: https://github.com/fitchmultz/pi-tool-duration/compare/v0.6.2...v0.6.3 ## 0.6.2 - 2026-10-31 - Automate reviewed npm patch releases (#38) ([f000cb5](https://github.com/fitchmultz/pi-tool-duration/commit/f000cb550f8e0cfc8b3ca5deafa2f01ffce2687a)) ### What's Changed - Automate reviewed npm patch releases by @fitchmultz in https://github.com/fitchmultz/pi-tool-duration/pull/38 **Full Changelog**: https://github.com/fitchmultz/pi-tool-duration/compare/v0.6.1...v0.6.2 ## [0.6.1] - 2026-10-31 ### Documentation - Rewrite the README around installation, examples, and configuration; add a tool-timing diagram and move technical and maintainer reference material into `docs/`. - Document the unpinned Git installation fallback. ### Tests - Check native Responses tool declarations on current official Pi and the maintained fork. ## [0.6.0] - 2026-10-01 ### Changed - Require Pi 1.0.0; qualify against the exact official 1.0.0 cohort and host TypeBox 1.3.27. - Cache timing and known absence per finalized result occurrence on the selected branch. Cold legacy recovery remains complete; warmed orphan requests no longer rewalk history, and appends process only their suffix. - Share one lookup for both compaction input arrays, skipping all history work for text-only compaction. - Keep same-ID/same-finalized-timestamp result occurrences separate, including their ordered request and compaction copies. - Use only official 1.0 APIs; retire metadata/checkpoint runtime branches dropped from the maintained fork. Preserve legacy checkpoint copies when reading old journals. ## [0.5.0] - 2026-10-01 ### Changed - Leave successful calls under 50 ms unannotated by default; their marker would read `0.0s`. Failed calls are still always annotated. Set `PI_TOOL_DURATION_THRESHOLD_MS=0` to annotate every result again. ### Removed - Retired fork-native async detach/resume timing and live continuation event integrations. Ordinary tool timing and saved timing markers remain supported. ## [0.4.1] - 2026-09-26 ### Fixed - Include the timing marker in results sent on the maintained fork's live WebSocket steering continuations, using its `live_tool_result` event. Recorded output is unchanged, and the live copy matches later requests. ## [0.4.0] - 2026-09-26 ### Fixed - Annotate every result when concurrent native async completions persist all of their timings before any result. Timings are matched by call ID up to the issuing assistant message. - Keep measured time for fork native async calls that detach and resume, including across restarts. The reported duration sums the observed active spans and excludes detached time. ### Changed - Require Node.js 24 or later and develop against Pi 0.87.1, TypeScript 7, and Node 24 types. - Regenerate the development lockfile from scratch on the public npm registry. ## [0.3.1] - 2026-09-22 ### Fixed - Use distinct native response IDs in compaction fixtures so unrelated tool calls cannot be mistaken for a single response. The runtime is unchanged from 0.3.0. ### Changed - Qualify current Pi hosts with focused, artifact-free compatibility checks. ## [0.3.0] - 2026-09-21 ### Changed - Annotate every tool result by default with `[host tool-call elapsed: Xs]`. Set `PI_TOOL_DURATION_THRESHOLD_MS=1000` to retain slow/failed-only reporting. - Require Pi 0.87.0 or later, using the same native extension APIs on official Pi and the maintained fork. - Resolve timings through a stateless backward ancestry lookup instead of rebuilding the complete historical timing map for every request. ### Fixed - Preserve prompt and tool updates in model context so duration annotations do not collapse the reusable request prefix. - Associate timing with the finalized result when a later extension replaces its timestamp, preserving existing saved timing records. ### Tests - Exercise native Responses serialization, dynamic prompt/tool prefixes, and finalized timestamps. - Verify actual timed results through disk restoration and maintained-fork checkpoints. ## [0.2.3] - 2026-09-18 ### Fixed - Preserve timing in native compaction input when long tool output is truncated, including split-turn summaries. Ordinary requests still append one duration marker, and recorded tool content and details remain unchanged. ## [0.2.2] - 2026-09-18 ### Fixed - Keep duration markers out of terminal tool output after reload, resume, and branch navigation by storing timing in hidden session entries and annotating only model-request copies. - Preserve model-visible timings across restoration, compaction summaries, and retained compaction history without changing tool content or details. ### Changed - Verify recorded output, reload, restored timing, branch isolation, and repeated model requests in regression tests. - Run clean-install checks on Node 22.19 and Node 24 in GitHub Actions. ## [0.2.1] - 2026-08-09 ### Fixed - Use Pi's error status instead of guessing from arbitrary result text or metadata, eliminating false duration markers on successful tools. - Append duration markers to finalized model-visible messages so the TUI keeps its single native timer, missing-content tools stay safe, blocked preflight failures are covered, and marker-like tool output is still measured. - Keep the long threshold flag readable in `pi --help` and declare the documented Pi 0.84.0 host floor in package metadata. - Keep the integration-test dependency lockfile installable from the public npm registry. ### Changed - Added real Pi CLI integration coverage over a local HTTP model transport for timing, flags, failures, parallel calls, lifecycle cleanup, and result preservation. ## [0.2.0] - 2026-08-06 ### Changed - Requires Pi 0.84.0 or later and refreshes the development lock against the released Pi 0.84.0 packages. - Verified the tool timing hooks and lifecycle cleanup against Pi 0.84.0's extension events, emitted types, and runtime implementation. ## [0.1.4] - 2026-07-16 - Refreshed the local Pi development lock and validation baseline to Pi 0.80.9; the unified model runtime/provider changes do not affect tool timing. ## [0.1.3] - 2026-07-14 ### Changed - Refreshed the local Pi development lock and validation baseline to Pi 0.80.7. ## [0.1.2] - 2026-07-11 ### Changed - Measure from Pi's `tool_execution_start` event so timing matches the full current execution lifecycle, including preflight hooks. - Clear abandoned timing state when the agent fully settles. ## [0.1.1] - 2026-07-03 ### Changed - Always append duration markers for tool results that report a non-zero exit code, even below the normal duration threshold. ## [0.1.0] - 2026-06-24 ### Added - Initial `pi-tool-duration` package. - Appends `[duration: Xs]` to model-visible Pi tool results when execution meets the configured threshold. - Supports `--tool-duration-threshold-ms` and `PI_TOOL_DURATION_THRESHOLD_MS` overrides.