Skip to content

Caudra 0.2 Preview 7

View Markdown~952 tokens

v0.2.0-preview.7 on GitHub

  • Context windows match the limits providers enforce, and can be set for built-in providers.
  • Global config files read secrets from files.
  • GPT reasoning on Copilot stays in one piece.
  • Copying in WSL reaches the Windows clipboard.
  • Only the plan menu approves a saved plan.
  • A session saves again after another Caudra carried it on.

Important

Keep {file:PATH} out of your config files until every Caudra process that reads them runs Preview 7. Older versions read it as literal text.

  • Copilot reasoning stays whole. GPT models on Copilot and Copilot Enterprise split their reasoning into a block per streamed word, sent it back to the model as plain reply text, and dropped the encrypted reasoning later turns need. Each reasoning summary is now one block, and later turns send the encrypted reasoning back as they do with OpenAI. Turns saved before the update keep their split blocks.
  • Context windows match the limits providers enforce. Copilot models now use Copilot's prompt limit as the window, so copilot-enterprise/gpt-6.1-sol runs at 272k instead of 400k and compacts before Copilot refuses the prompt. OpenCode models with a separate input limit, such as the GPT-5 models, now run at their full input limit instead of compacting early. On any model whose reply shares the window, compaction now leaves room for the reply, up to half the window, so a model like claude-haiku-4-5, with 64k of a 200k window, is no longer refused near the limit. /context and the status bar show the new point.
  • Set a built-in model's window in providers.toml. A [[SLUG.models]] entry or a [SLUG.model_defaults] table now sets context_window and max_output_tokens for a built-in provider such as anthropic, ahead of what the provider lists. Caudra warns about any other model key there, since built-in providers ignore it. See Model limits on a built-in provider.
  • Values from files. Beside {env:NAME}, a string value in the global caudra.toml, permissions.toml, mcp.toml, providers.toml, or workcell.toml can now hold {file:PATH}. Caudra puts the file's text there, trimmed at both ends, so a secret kept in a file stays out of the config. A relative path starts at the config file's directory, and a path can hold {env:NAME}, such as {file:{env:CREDENTIALS_DIRECTORY}/token}. The file must be a UTF-8 regular file of at most 1 MiB. Project files under .caudra/ cannot use references. See Values from files.
  • Copying works in WSL. Copied text often stayed in the WSLg clipboard and never reached Windows. Every copy in WSL now also goes through OSC 52, which Windows Terminal and Orca pass to the Windows clipboard. In Orca, this needs Allow TUI Clipboard Writes (OSC 52), which older Orca releases leave off.
  • A saved plan goes to the plan menu. With the question-tool nudge on, the model could ask whether to implement a saved plan, and that answer could not leave plan mode. Once the plan is saved, the nudge now stays off until your next message and the model ends with a summary. Approve the plan from the plan menu as before.
  • Reopening a session another Caudra carried on saves again. If you left a session with Ctrl+X n or /new, carried it on in another Caudra, and reopened it in the first, every save failed with "session … was modified concurrently", and no other Caudra could open it until you quit. Deleting it from the session picker failed the same way. Both now work. The session on disk was never damaged.

Website privacy and analytics