Skip to content

TUI Commands

The TUI supports vim-style : commands for runtime configuration and control. Type : at the beginning of a line to enter command mode.

CommandDescription
:set option=valueSet configuration option
:set option?Query current value
:set option??Show modification history
:set option!Toggle boolean option
:set option&Reset to default
:setShow modified options
:set allShow all options
:modelOpen model picker
:model <name>Switch to model
:session listList sessions
:session load <id>Load session
:quit / :qExit chat
:helpShow help

Crucible’s :set command follows Vim conventions for runtime configuration.

:set option=value # Set string/number option
:set option:value # Alternative syntax
:set option value # Space-separated (if unambiguous)

Examples:

:set model=claude-3-5-sonnet
:set thinkingbudget=8000
:set temperature=0.7
:set option # Enable boolean option
:set nooption # Disable boolean option (prefix with 'no')
:set option! # Toggle option
:set invoption # Toggle option (alternative)

Examples:

:set thinking # Enable thinking display
:set nothinking # Disable thinking display
:set verbose! # Toggle verbose mode
:set option? # Show current value
:set option?? # Show modification history
:set # Show all modified options
:set all # Show all options with values
:set option& # Reset to default value
:set option^ # Pop one modification (undo last change)
OptionTypeDescription
modelstringCurrent LLM model (e.g., claude-3-5-sonnet, gpt-4o)
providerstringLLM provider (ollama, openai, anthropic)
OptionTypeDescription
thinkingboolShow thinking/reasoning tokens in UI
thinkingbudgetnumber/presetToken budget for extended thinking

Thinking Budget Presets:

PresetTokensDescription
off0Disable extended thinking
minimal512Brief reasoning
low1024Light reasoning
medium4096Moderate reasoning
high8192Thorough reasoning
maxunlimitedMaximum reasoning

Examples:

:set thinkingbudget=4096 # Set exact token count
:set thinkingbudget=high # Use preset
:set thinkingbudget=off # Disable thinking
OptionTypeDescription
themestringSyntax highlighting theme
verboseboolVerbose output mode
OptionTypeDescription
temperaturefloatResponse randomness (0.0 - 2.0)
maxtokensnumberMaximum response tokens

Switch models at runtime:

:model # Open model picker popup
:model <name> # Switch directly to model

The model picker shows available models from your configured provider. Navigate with arrow keys, select with Enter.

Examples:

:model claude-3-5-sonnet
:model gpt-4o
:model llama3.2

Model changes persist for the session and sync to the daemon (if using daemon mode).

Manage chat sessions:

:session list # Show available sessions
:session load <id> # Resume existing session
:session new # Start new session

Sessions auto-save and can be resumed across TUI restarts.

:quit # Exit chat (alias: :q)
:help # Show help
:clear # Clear conversation (start fresh)
:palette # Open command palette

The :set command modifies a runtime overlay on top of your base configuration:

┌─────────────────────────────┐
│ :set commands (runtime) │ ← Highest priority
├─────────────────────────────┤
│ Environment variables │
├─────────────────────────────┤
│ ~/.config/crucible/ │
│ config.toml (user) │
├─────────────────────────────┤
│ Built-in defaults │ ← Lowest priority
└─────────────────────────────┘

Runtime changes do not persist to config files. They last for the current session only.

Use :set option?? to see where a value came from:

:set thinkingbudget??
# Output:
# thinkingbudget = 8192
# [Command] 8192 (2025-01-20 14:30:00)
# [File] 4096 (base config)

Some options have short aliases:

ShortcutFull Path
modelllm.providers.{provider}.default_model
thinking(virtual, TUI-only)
thinkingbudgetllm.thinking_budget
themecli.highlighting.theme
verbosecli.verbose
:model gpt-4o
:set thinking
:set thinkingbudget=high
:set model?
:set thinkingbudget?
:set thinkingbudget&
:set temperature&
:set all # See everything
:set # See what you changed
:set model?? # See modification history