Multilingual Support Squad Template

Overview

This page covers a structured approach to multilingual support built as a Squad: callers explicitly choose their language up front, then get routed to a dedicated assistant built and tuned for that language. This is a different approach from the single-assistant automatic language detection covered in Multilingual Assistants — see the comparison at the end of this page for when to use each.

The squad is built around a short entrance flow followed by dedicated per-language assistants:

1. Entrance flow – the first assistant presents a brief language selection, for example "Press or say 1 for English, 2 for Spanish, 3 for French."
2. Routing – based on the caller's selection, the call is handed off to the matching language assistant.
3. Dedicated assistant – each language assistant has its own tuned system prompt and a voice appropriate to that language and region.

Each squad member is a language-specific assistant with its own prompt and voice:

{
  "members": [
    {
      "assistant": {
        "name": "English Support",
        "model": {
          "provider": "openai",
          "model": "gpt-4o",
          "messages": [
            { "role": "system", "content": "English support. Direct, friendly, professional." }
          ]
        },
        "voice": { "provider": "azure", "voiceId": "en-US-AriaNeural" },
        "firstMessage": "Hello! How can I help you today?",
        "firstMessageMode": "assistant-speaks-first"
      }
    },
    {
      "assistant": {
        "name": "Soporte Espanol",
        "model": {
          "provider": "openai",
          "model": "gpt-4o",
          "messages": [
            { "role": "system", "content": "Soporte en espanol. Calido y respetuoso; usa 'usted' inicialmente." }
          ]
        },
        "voice": { "provider": "azure", "voiceId": "es-ES-ElviraNeural" }
      }
    },
    {
      "assistant": {
        "name": "Support Francais",
        "model": {
          "provider": "openai",
          "model": "gpt-4o",
          "messages": [
            { "role": "system", "content": "Support francais. Poli, courtois et formel." }
          ]
        },
        "voice": { "provider": "azure", "voiceId": "fr-FR-DeniseNeural" }
      }
    }
  ]
}

When the entrance assistant hands off to a language-specific assistant, the squad's handoff mechanism carries the conversation's context forward, so the caller doesn't need to repeat information already given during the language-selection step.

This structured Squad approach differs from the single-assistant approach described in Multilingual Assistants, which uses automatic language detection instead of an explicit selection step.

Choose the Squad template when you want the strongest per-language tuning and are comfortable with a brief selection step. Choose the single-assistant approach from Multilingual Assistants when you want callers to speak naturally in any supported language without choosing one first.