Create Assistant

Create a new assistant configuration.

• URL: /assistant/create
• Method: POST
• Headers: Authorization: Bearer <your_api_key>
• Content-Type: application/json

Request Body (Common Fields)

Field	Type	Required	Description
assistant_name	string	Yes	Assistant name (1-100 chars).
assistant_description	string	Yes	Assistant description.
assistant_prompt	string	Yes	System prompt.
assistant_llm_mode	string	No	LLM mode: pipeline or realtime. Default: pipeline.
assistant_start_instruction	string	No	Opening response text. Used when assistant_interaction_config.speaks_first=true (max 200 chars).
assistant_interaction_config	object	No	Interaction settings (see below).
assistant_end_call_enabled	boolean	No	Enables built-in end-call tool. Default: false.
assistant_end_call_trigger_phrase	string	Conditional	Required if assistant_end_call_enabled=true.
assistant_end_call_agent_message	string	Conditional	Required if assistant_end_call_enabled=true.
assistant_end_call_url	string	No	Webhook URL for call-ended payload.

---

Mode Configuration

Pipeline Realtime

Pipeline mode uses OpenAI Realtime for STT + LLM, with a separate TTS provider for audio output. If assistant_interaction_config.speaks_first=true, the opening response is spoken at session start.

Required fields

Field	Type	Required	Description
assistant_tts_model	string	Yes	One of cartesia, sarvam, elevenlabs, mistral.
assistant_tts_config	object	Yes	TTS config for the selected provider (see tabs below).

TTS provider configuration

CartesiaSarvamElevenLabsMistral

Field	Type	Required	Description
voice_id	string	Yes	Cartesia voice ID.
api_key	string	No	Optional Cartesia API key.

Field	Type	Required	Description
speaker	string	Yes	Sarvam speaker identifier.
target_language_code	string	No	BCP-47 code. Default: bn-IN.
api_key	string	No	Optional Sarvam API key.

Field	Type	Required	Description
voice_id	string	Yes	ElevenLabs voice ID.
api_key	string	No	Optional ElevenLabs API key.

Field	Type	Required	Description
voice_id	string	Yes	Mistral voice ID.
api_key	string	No	Optional Mistral API key.

Example request

curl -X POST "https://api-livekit-vyom.indusnettechnologies.com/assistant/create" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <your_api_key>" \
  -d '{
    "assistant_name": "Support Bot",
    "assistant_description": "First line support",
    "assistant_prompt": "You are a helpful customer support agent.",
    "assistant_llm_mode": "pipeline",
    "assistant_tts_model": "cartesia",
    "assistant_tts_config": {
      "voice_id": "a167e0f3-df7e-4277-976b-be2f952fa275"
    },
    "assistant_interaction_config": {
      "speaks_first": true,
      "filler_words": true,
      "silence_reprompts": true,
      "silence_reprompt_interval": 10.0,
      "silence_max_reprompts": 2
    }
  }'

Realtime mode uses a single model (e.g. Gemini Live API) that handles STT, LLM, and TTS in one stream. If assistant_interaction_config.speaks_first=true, the opening response is sent at session start through the realtime conversation path.

Filler words are not available in realtime mode

Since there is no external TTS, filler_words is automatically disabled even if set to true.

Required fields

Field	Type	Required	Description
assistant_llm_config	object	Yes	Realtime provider configuration (see table below).

Realtime LLM config (assistant_llm_config)

Field	Type	Required	Description
provider	string	No	Realtime provider. Current value: gemini.
model	string	No	Gemini realtime model. Default: gemini-3.1-flash-live-preview.
voice	string	No	Gemini voice. Default: Puck.
api_key	string	No	Optional per-assistant Google key. Falls back to system GOOGLE_API_KEY.

Example request

curl -X POST "https://api-livekit-vyom.indusnettechnologies.com/assistant/create" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer <your_api_key>" \
  -d '{
    "assistant_name": "Gemini Assistant",
    "assistant_description": "Realtime voice assistant",
    "assistant_prompt": "You are a helpful assistant.",
    "assistant_llm_mode": "realtime",
    "assistant_llm_config": {
      "provider": "gemini",
      "model": "gemini-3.1-flash-live-preview",
      "voice": "Puck"
    }
  }'

---

Interaction Configuration

Field	Type	Required	Description
speaks_first	boolean	No	If true (default), assistant sends an opening response first in both pipeline and realtime modes.
filler_words	boolean	No	Enables filler words while user is speaking. Pipeline mode only.
silence_reprompts	boolean	No	Enables reprompts during prolonged user silence.
silence_reprompt_interval	number	No	Reprompt interval in seconds (1.0-60.0). Default: 10.0.
silence_max_reprompts	number	No	Maximum reprompts before ending session (0-5). Default: 2.

Response Schema

Field	Type	Description
success	boolean	Operation status.
message	string	Human-readable message.
data.assistant_id	string	Created assistant UUID.
data.assistant_name	string	Created assistant name.

Example Response

{
  "success": true,
  "message": "Assistant created successfully",
  "data": {
    "assistant_id": "550e8400-e29b-41d4-a716-446655440000",
    "assistant_name": "Support Bot"
  }
}

HTTP Status Codes

Code	Description
200	Assistant created successfully.
400	Validation or payload mismatch error.
401	Unauthorized.
500	Internal server error.