Configuration

Talkform is driven by AudioformConfig.

Shape

type AudioformConfig = {
  id: string;
  title: string;
  description?: string;
  instructions?: string;
  fields: AudioformField[];
  theme?: {
    accent?: string;
    surface?: string;
    panel?: string;
  };
  realtime?: {
    model?: string;
    voice?: string;
  };
  output?: {
    formats?: Array<"json" | "markdown">;
    webhookUrl?: string;
  };
};

Field types

• text
• long_text
• single_select
• multi_select
• number
• rating
• url
• file_ref

Every field includes:

• id
• label
• type
• required
• promptTitle
• promptDetail
• options?
• validation?
• agentHint?

Design guidance

• Keep prompts short and concrete.
• Use single_select and multi_select whenever the downstream system expects enumerated values.
• Put product-specific logic downstream of the exported JSON instead of inside the Talkform core.