Skip to main content
GET
/
config
/
pipeline-options
List Pipeline Options
curl --request GET \
  --url https://voice-livekit.studio.lyzr.ai/v1/config/pipeline-options \
  --header 'x-api-key: <api-key>'
{
  "stt": [
    {
      "providerId": "deepgram",
      "displayName": "Deepgram",
      "models": [
        {
          "id": "deepgram/nova-3:en",
          "name": "Deepgram Nova-3 (EN)",
          "languages": [
            "<string>"
          ]
        }
      ]
    }
  ],
  "tts": [
    {
      "providerId": "elevenlabs",
      "displayName": "ElevenLabs",
      "models": [
        {
          "id": "elevenlabs/eleven_turbo_v2_5",
          "name": "ElevenLabs Eleven Turbo v2.5",
          "defaultVoiceId": "Xb7hH8MSUJpSbSDYk0k2",
          "languages": [
            "<string>"
          ]
        }
      ]
    }
  ],
  "llm": [
    {
      "providerId": "openai",
      "displayName": "OpenAI",
      "models": [
        {
          "id": "openai/gpt-4o",
          "name": "GPT-4o"
        }
      ]
    }
  ]
}

Documentation Index

Fetch the complete documentation index at: https://docs.lyzr.ai/llms.txt

Use this file to discover all available pages before exploring further.

Discover the complete catalog of models available for building a modular Voice Agent pipeline. While “Realtime” models (like gpt-realtime or gemini-native-audio) handle listening, thinking, and speaking in one step, a Pipeline agent strings together three separate specialized models. This endpoint gives you the exact IDs needed to configure that 3-step chain.
Authentication Required: You must include your API key in the x-api-key header to authenticate this request.

​
Understanding the Pipeline Architecture

To create a pipeline agent, you must configure three components in your POST /agents request. The response from this endpoint provides the available options for each:

​
1. stt (Speech-to-Text)

This is the model that transcribes the user’s spoken audio into text.
  • Available Providers: Deepgram, AssemblyAI, Cartesia, Sarvam.
  • Key Detail: Pay close attention to the languages array for each STT model to ensure it supports your target demographic.

​
2. llm (Large Language Model)

This is the “brain” of the agent. It takes the text from the STT model, processes it against your system prompt, and generates a text response.
  • Available Providers: OpenAI (GPT-4o, GPT-5 series), Google (Gemini Flash/Pro series), DeepSeek, MoonshotAI.

​
3. tts (Text-to-Speech)

This model takes the text generated by the LLM and synthesizes it into spoken audio for the user to hear.
  • Available Providers: ElevenLabs, Cartesia, Deepgram (Aura), Inworld, Rime, Sarvam.
  • Key Detail: The response includes a defaultVoiceId for each TTS model, which you can use as a fallback if you aren’t rendering a full list of custom voice clones.

Authorizations

​
x-api-key
string
header
required

Response

200 - application/json

Pipeline options retrieved successfully.

​
stt
object[]

Supported Speech-to-Text (transcription) providers and models.

​
tts
object[]

Supported Text-to-Speech (voice generation) providers and models.

​
llm
object[]

Supported text-based Large Language Models used for reasoning.