mirror of
https://github.com/hanzoai/dev.git
synced 2026-10-17 15:34:54 +00:00
The docs read as OpenAI's Codex CLI: a getting-started page titled for it, `codex --enable skills` and `code sandbox macos` for commands that do not exist, defaults given as gpt-5.1-codex-max/openai when they are zen5-coder/hanzo, a Homebrew page describing a formula we do not publish. Removed three tui-stream-chunking notes and the execpolicy page: they document modules and a subcommand that live only in the vendored upstream copy. There is no execpolicy subcommand and no .codexpolicy loader in code-rs; the page told users to run something the binary does not have. Rewrote tui-alternate-screen against the source. It described an auto/always/never enum and a --no-alt-screen flag; the shipped behavior is a tui.alternate_screen bool with Ctrl+T, which replays the transcript into scrollback on the way out — a better answer to Zellij than the one written down. Genuine openai-provider documentation stays.
3.1 KiB
3.1 KiB
TUI Settings Overlay
Full-screen settings panel for Hanzo Dev’s TUI. Use it to change models, themes, Auto Drive defaults, agents, notifications, and more without leaving the chat.
Open & navigate
/settingsopens the overview;/settings <section>jumps directly (section names below)./auto settingsand/updateroute into their sections.- Keys:
↑/↓orj/kto move,Tab/Shift+Tabto cycle sections,Home/Endjump list ends. Enter opens/activates; Esc closes the current section, Esc again closes the overlay.?toggles inline help. Paste is forwarded to the active section when allowed. - Overlay is modal: chat input is blocked while it is visible. It remembers the last active section on reopen (
pending_settings_return).
Persistence
- Changes write to
CODE_HOME/config.tomlwhen available; if that directory is missing you’ll see a warning and the changes remain session-only. - Access mode can be stored per workspace; other settings apply globally unless your config file overrides them per project.
- Agent and MCP edits also live in the same config directory.
Sections
- Model: pick the default chat model and reasoning effort.
- Theme: choose a theme and spinner; applies immediately.
- Updates: view upgrade channel/status.
/updateopens here before running installers. - Agents: see built-in/custom agents, enable/disable, force read-only, add per-agent instructions. Open the Subagent editor to configure
/plan//solve//codeor custom slash commands. - Prompts: edit saved prompt snippets.
- Auto Drive: set review/agents/QA/cross-check toggles, continue mode (manual/immediate/ten-seconds/sixty-seconds), model override, or “use chat model.” Updates apply to active runs.
- Review: choose a review model (or reuse chat), toggle auto-resolve, and set the max auto-resolve attempts.
- Planning: pick model/effort for planning turns or reuse the chat model.
- Validation: toggle validation groups and tools; see install status; trigger install help.
- Limits: read-only view of rate limits and context/auto-compact usage.
- Chrome: shown when browser attach fails; choose to retry, use a temp profile, switch to the internal browser, or cancel.
- MCP: enable/disable MCP servers.
- Notifications: toggle all or set filters.
/notifications on|off|statusalso lands here.
Overlay lifecycle
- All keys route through
handle_settings_keywhile open; composer/history ignore input until closed. - Help overlay (
?) closes before the main overlay when Esc is pressed. - Sections mark completion via their content structs; the overlay closes when a section reports
is_complete(e.g., Chrome option chosen).
Scope reminders
- Global defaults live in
CODE_HOME/config.toml. - Workspace overrides are honored where setters accept
cwd(access mode) or when project-level config files exist. The UI always renders merged effective values. - Agent commands and MCP servers are stored under
CODE_HOMEand apply to all workspaces unless overridden by project config.
Commands
/settings [section]/auto settings/update(or/update settings)/notifications [on|off|status]