Skip to content

0.179.0

Choose a tag to compare

@vinhnx vinhnx released this 16 Oct 14:42
· 912 commits to main since this release

What's Changed

0.179.0 - 2026-10-16

Highlights

Bug Fixes

  • Invalidate cached exec-slice reads on mutation; drop reuse-stub re-attachment (e133f12)
  • Keep mid-turn prompt verbatim across model switch (5472fd7)

Documentation

  • Document exec-slice reuse and history-verbatim reference stubs (695eb2e)

Features

  • Record bounded raw payload when structured JSON fails to parse (ca950fa)
  • Update cloud model catalog (cec48a8)
  • Reuse bounded shell slice reads across turns via shared exec read-target parser (088c025)
  • Serve bounded reference stub when reuse response duplicates history content (0f4c741)

Other Changes

Other

  • Refactor LLM error handling and enhance Anthropic speed configuration (8e40555)
  • Add eval plan (a5f6304)
  • Add SWE bench eval plan (f6ce0bb)

Refactors

  • Add Provider::name_matches as single canonical name matcher (54aa9e5)
  • Consolidate local_server base-URL resolution; lock z.ai legacy-env precedence (dda8281)

Full Changelog: 0.178.0...0.179.0