Skip to content

Releases

v0.0.140

Choose a tag to compare

@hsnchebi hsnchebi released this 26 Oct 02:40
1c92f66

[0.0.140] - 2026-10-26

Added

  • Claude Sonnet 5.5 and Claude Haiku 5.5 on the Anthropic and Claude Code providers; Claude Opus 5.5, Fable 5.1, Sonnet 5.5 and Haiku 5.5 on Bedrock and Vertex.
  • GPT-6.1 Sol on the OpenAI and Codex providers.
  • Model routing: send agents (Quick, Supervisor, Subagent, custom agents) to specific profiles via /profile → Model routing or multi profile route; --route agent=profile overrides routing for a single -p run. The active routing shows in the footer, /stats and the settings list.
  • Markdown tables are drawn with box-drawing borders.
  • Update Todos lines show progress and the next step.
  • The @ file menu is available when answering a question.
  • Cmd+Left/Right jump to the start/end of a line, Cmd+Up/Down to the start/end of the input.
  • The display stays awake while an agent run is active (--no-keep-awake or MULTI_KEEP_AWAKE=0 to turn it off).
  • Large pastes land at once, including on terminals without bracketed paste.

Changed

  • Emoji and icons are replaced only on the classic Windows console, which can't draw them; other Windows and WSL terminals just pad them. MULTI_EMOJI=keep|pad|icons|ascii overrides the detection.
  • The CLI embeds Node 24.22.1 (October 2026 security release).

Fixed

  • A refused Anthropic, Vertex or MiniMax reply shows the refusal reason and ends the run instead of retrying.
  • Bedrock no longer sends temperature to Claude models that reject it.
  • Exhausted plan quotas and account balances are no longer retried.
  • Search keeps its results when some paths are unreadable.
  • Esc closes the @ menu instead of clearing the input.
  • The highlighted choice stays visible on short terminals.
  • The banner stays above a prompt typed while it is animating.

v0.0.139

Choose a tag to compare

@hsnchebi hsnchebi released this 01 Oct 20:43
1c92f66

[0.0.139] - 2026-10-01

Added

  • /continue resumes the interrupted run of a session; multi --resume <id> --continue does the same without the interactive UI.
  • /copy copies the last reply as markdown.
  • Profile setup and editing offer optional thinking, reasoning effort and verbosity steps; clearing one resets it to the model default.
  • Worker views show the delegated task at the top and a jump-to-latest control, and each view keeps its scroll position when switching between main and worker views.
  • Emoji are replaced or padded on Windows and WSL terminals so they render at the right width.

Fixed

  • Markdown tables wrap to fit the terminal width, including in replayed sessions.
  • Wrapped text keeps its styling on every row.
  • Errors now show in the TUI instead of only in the log.

v0.0.137

Choose a tag to compare

@hsnchebi hsnchebi released this 29 Sep 21:59
1c92f66

[0.0.137] - 2026-09-29

Added

  • Up/Down move between lines of a multi-line input; history is only recalled at the first or last line.
  • Answers to ask-question prompts can contain newlines.

Fixed

  • Empty reasoning blocks now show as "Reasoning…" then "Reasoned for Ns", and the timer no longer runs on into the next block.
  • Long input is capped to a window that follows the caret, so its bottom stays visible.
  • Long ask-question options wrap under their text.
  • Repeated rewinds now anchor to the correct message.
  • Plugin-host logs go to the log file instead of the console.

v0.0.135

Choose a tag to compare

@hsnchebi hsnchebi released this 24 Sep 14:42
1c92f66

[0.0.135] - 2026-09-24

Added

  • Claude Opus 5.5 on the Anthropic and Claude Code providers. Now the default model.
  • Claude Fable 5.1 on the Anthropic and Claude Code providers.

Removed

  • Retired claude-opus-4-1 model (retired by Anthropic on August 5, 2026).

v0.0.131

Choose a tag to compare

@hsnchebi hsnchebi released this 04 Sep 17:44
1c92f66

[0.0.131] - 2026-09-04

Initial release.

latest

Choose a tag to compare

@hsnchebi hsnchebi released this 04 Sep 17:45
1c92f66

Pointer to the current release. Not a release itself.