deepagents-code==0.2.7

Features

• Added shell completion scripts via dcode completion (#7264).

Bug Fixes

• Fixed detection of editor launchers such as code.cmd on Windows (#7642).
• Improved dcode completion output and setup steps, and fixed Bash completion for --flag=value arguments (#7658, #7656).
• Preserved compaction feedback and offload usage when summaries do not stream (#7644).

Performance Improvements

• Reduced Git attribution overhead by halving probes, backing off after failures, and scanning the worktree only once per evidence probe. Added an option to disable attribution probes (#7661, #7660).

Full changelog

---

Thanks to our community contributors: Eric (@kuchaguangjie), Archit Shukla (@Arsh624) (LinkedIn)

Special thanks to everyone who reported the issues addressed in this release: Eric (@kuchaguangjie) (#7238), Archit Shukla (@Arsh624) (#7620)

Internal maintainers: Mason Daugherty (@mdrxy)

Released by: Mason Daugherty (@mdrxy)

Git log since deepagents-code==0.2.6

This commit history includes changes to this package. The most recent commit is at the top; older commits follow below.

• 6a8d452 release(deepagents-code): 0.2.7 (#7654)
• 7ffdf97 perf(code): halve git attribution probes, back off on failures, add opt-out (#7661)
• 172814d perf(code): scan the worktree once per git evidence probe (#7660)
• c41e786 fix(code): complete --flag=value in bash (#7656)
• 277c8b5 fix(code): harden dcode completion output and setup steps (#7658)
• 6217497 test(code): cover nested levels and fish in completion tests (#7657)
• 41b8cdb feat(code): add shell completion scripts via dcode completion (#7264)
• de00d21 chore(deps): raise dependency minimums (#7653)
• 0edc892 fix(code): keep compaction feedback and offload usage when summaries don't stream (#7644)
• 1ea2055 fix(code): find editor launchers like code.cmd on Windows (#7642)
• 079d590 chore(deps): raise dependency minimums (#7649)