Starstruk Logo MarkStarstruk

v0.46.0

rtk-ai/rtkv0.46.0Aug 26, 2026by rtk-release-bot[bot]

AI Summary

This release focuses on significant improvements to the 'find' command, including grammar dispatch and output compression, alongside various bug fixes for Git, Windows compatibility, and stream processing. It also addresses CI/CD benchmarking issues and test reliability.

Key Highlights

  • Enhanced 'find' command with grammar dispatch and output compression for unmodeled predicates.
  • Improved Windows compatibility by fixing console output decoding and drive-letter sanitization.
  • Fixed Git detection issues regarding patch flags and preserved patch output from log commands.
  • Enhanced stream processing to handle invalid UTF-8 by decoding lossily instead of dropping lines.
  • Resolved CI/CD benchmarks to prevent deletion of tracked scripts and handle negative cases.

New Features

  • Dispatch on find's grammar and compress find output for unmodeled predicates.
  • Tee tail hint when rtk imposes the result cap.

Full Release Notes

## [0.46.0](https://github.com/rtk-ai/rtk/compare/v0.45.0...v0.46.0) (2026-08-26)


### Features

- find: dispatch on find's grammar; compress find output for unmodeled predicates ([#3603](https://github.com/rtk-ai/rtk/pull/3603))
- find: tee tail hint when rtk imposes the result cap ([#3603](https://github.com/rtk-ai/rtk/pull/3603))

### Bug Fixes

- find: never-worse guard, recovery hint, and dispatch on find's grammar ([#3603](https://github.com/rtk-ai/rtk/pull/3603))
- git: don't misdetect a value-taking option's argument as a patch flag ([#3575](https://github.com/rtk-ai/rtk/pull/3575))
- cicd: stop benchmark.sh deleting the tracked scripts/benchmark harness ([#3595](https://github.com/rtk-ai/rtk/pull/3595))
- tee: hash long recovery-file slugs to prevent collisions and shorten hints ([#3266](https://github.com/rtk-ai/rtk/pull/3266))
- benchmark: avoid negative curl/cargo cases that fail the benchmark job ([#3430](https://github.com/rtk-ai/rtk/pull/3430))
- test: accept both Ask and Allow verdicts in rewrite tests ([#3147](https://github.com/rtk-ai/rtk/pull/3147)) — Closes [#3146](https://github.com/rtk-ai/rtk/issues/3146)
- core: decode process output using Windows console code page ([#2717](https://github.com/rtk-ai/rtk/pull/2717)) — Closes [#2452](https://github.com/rtk-ai/rtk/issues/2452)
- git: preserve patch output from log commands ([#2951](https://github.com/rtk-ai/rtk/pull/2951)) — Closes [#2944](https://github.com/rtk-ai/rtk/issues/2944)
- discover: sanitize drive-letter colon so Windows discover finds sessions ([#2952](https://github.com/rtk-ai/rtk/pull/2952)) — Closes [#2919](https://github.com/rtk-ai/rtk/issues/2919)
- stream: decode lossily instead of dropping lines on invalid UTF-8 ([#2997](https://github.com/rtk-ai/rtk/pull/2997)) — Closes [#2994](https://github.com/rtk-ai/rtk/issues/2994)

### Other

- test(find): use the platform temp dir instead of /tmp ([#3717](https://github.com/rtk-ai/rtk/pull/3717))