Skip to content

Configuration

Customize Dial Code behavior with settings files.

Settings Files

Settings can be defined at multiple levels:

Location Scope
~/.dial/settings.json Global (all projects)
.dial/settings.json Project (current directory)

Project settings override global settings.

Quick Start

Create a settings file:

mkdir -p ~/.dial
echo '{}' > ~/.dial/settings.json

Or use the settings dialog:

/settings

Common Settings

Session Token Limit

Control conversation length:

{
  "sessionTokenLimit": 32000
}

Execution Mode

Set default execution behavior:

{
  "executionMode": {
    "default": "auto",
    "autoSelect": true,
    "enableEscalation": true
  }
}

Model Selection

Choose your default model:

{
  "model": "qwen3-coder-plus"
}

Theme

Set the UI theme:

{
  "theme": "dracula"
}

Full Settings Reference

General

{
  "vim": false,
  "language": "auto",
  "autoUpdate": true,
  "checkpointing": true
}
Setting Type Default Description
vim boolean false Enable vim mode
language string "auto" UI language (auto, en, zh)
autoUpdate boolean true Check for updates
checkpointing boolean true Enable session recovery

UI

{
  "theme": "default",
  "hideTitleBar": false,
  "hideContextSummary": false,
  "hideFooter": false,
  "showLineNumbers": true
}
Setting Type Default Description
theme string "default" Color theme
hideTitleBar boolean false Hide title bar
hideContextSummary boolean false Hide context display
hideFooter boolean false Hide footer
showLineNumbers boolean true Show line numbers in code

Model

{
  "model": "qwen3-coder-plus",
  "sessionTokenLimit": 32000,
  "maxSessionTurns": -1,
  "generationConfig": {
    "timeout": 120000,
    "maxRetries": 3
  }
}
Setting Type Default Description
model string varies Model to use
sessionTokenLimit number 32000 Max tokens per session
maxSessionTurns number -1 Max turns (-1 = unlimited)

Tools

{
  "sandbox": false,
  "toolOutputTruncation": {
    "threshold": 10000,
    "maxLines": 500
  }
}
Setting Type Default Description
sandbox boolean false Enable sandboxed execution
toolOutputTruncation object varies Truncate large outputs

MCP Servers

{
  "mcpServers": {
    "server-name": {
      "command": "mcp-server-command",
      "args": ["--arg1", "value"],
      "env": {
        "API_KEY": "..."
      }
    }
  }
}

See MCP Integration for details.

Authentication

{
  "authType": "qwen-oauth"
}
Value Description
qwen-oauth Qwen OAuth (default)
openai-compatible API key authentication

Environment Variables

Settings can also be set via environment:

Variable Setting
OPENAI_API_KEY API key
OPENAI_BASE_URL API endpoint
OPENAI_MODEL Model name
DIAL_SANDBOX Sandbox mode

Environment variables take precedence over settings files.

Project-Specific Settings

Create .dial/settings.json in your project root:

{
  "model": "gpt-4",
  "sessionTokenLimit": 16000,
  "sandbox": true
}

This only affects sessions started in that directory.

Experimental Features

Enable experimental features:

{
  "experimental": {
    "visionModelPreview": true,
    "vlmSwitchMode": "once"
  }
}

Warning

Experimental features may change or be removed.

Resetting Settings

Delete settings files to reset:

# Reset global settings
rm ~/.dial/settings.json

# Reset project settings
rm .dial/settings.json

Next Steps