Model Settings

View as MarkdownOpen in Claude

Model Settings control how the AI behaves — the model powering responses, language handling, and formatting preferences.

Location: Left Sidebar → Agent Settings → Model tab

Model settings
Model tab for Single Prompt agents

AI Model

Choose the LLM powering your agent and its primary language.

SettingDescription
LLM ModelThe AI model (Electron, GPT-4o, etc.)
LanguagePrimary language for responses

You can also set the model in the Prompt Section dropdown at the top of the editor.


Speech Formatting

When enabled (default: ON), the system automatically formats transcripts for readability — adding punctuation, paragraphs, and proper formatting for dates, times, and numbers.


Language Switching

Enable your agent to switch languages mid-conversation based on what the caller speaks (default: ON).

Advanced Settings

When Language Switching is enabled, you can fine-tune detection:

SettingRangeDefaultWhat it does
Minimum Words for Detection1-102How many words before considering a switch
Strong Signal Threshold0-10.7Confidence level for immediate switch
Weak Signal Threshold0-10.3Confidence level for tentative detection
Consecutive Weak Signals1-82How many weak signals needed to switch

Understanding the thresholds:

  • Strong Signal: Very confident the caller switched → switches immediately
  • Weak Signal: Somewhat confident → waits for more evidence
  • Higher thresholds = More certain before switching (fewer false switches)
  • Lower thresholds = Quicker to switch (more responsive)

For most cases, the defaults work well. Adjust only if you’re seeing unwanted switching behavior.