Skip to main content
Version: 2.0

Vectara and Speechmatics

Speechmatics provides real-time speech-to-text transcription with low latency in more than 55 languages. By combining Vectara's RAG capabilities and the LiveKit voice agent framework, you can build conversational AI assistants with enterprise-grade accuracy.

Integration benefits​

  • Enables real-time voice interactions with your Vectara corpora.
  • Provides low latency (less than 1 second) speech-to-text transcription in 55+ languages.
  • Leverages LiveKit's open-source agent framework for seamless voice agent orchestration.
  • Supports flexible deployment options: cloud, on-premises, or on-device.

Architecture​

The integration combines three key components:

  1. Speechmatics - Real-time speech-to-text transcription and text-to-speech (TTS) for agent response.
  2. Vectara - Knowledge retrieval and RAG-powered responses.
  3. LiveKit - Voice agent orchestration and audio streaming.

Getting started​

To build your own voice agent with Vectara and Speechmatics:

  1. Set up LiveKit - Follow the LiveKit Agents quickstart.
  2. Configure Speechmatics - Add your Speechmatics API key for real-time STT.
  3. Connect Vectara - Integrate Vectara's query API for knowledge retrieval.
  4. Add text-to-speech (TTS) - Choose a text-to-speech provider such as Speechmatics or ElevenLabs.