Skip to main content
This page covers the key configuration options for Burki Voice AI assistants and integrations.

​
Provider Configuration

Burki supports two configuration modes for AI providers:

​
Burki Cloud (Managed)

When using Burki Cloud, no API keys are required. Simply:
  1. Select Burki Cloud mode in your assistant settings
  2. Choose your preferred providers and models
  3. Costs are automatically billed to your wallet

​
BYO (Bring Your Own)

When using your own API keys, configure them in Settings → Provider Keys:

​
Assistant Configuration

Each assistant can be configured with:

​
Basic Settings

  • Name — Display name for the assistant
  • Description — Internal notes about the assistant’s purpose
  • Greeting Message — First message spoken when a call connects

​
AI Configuration

  • LLM Provider — OpenAI, Azure OpenAI, Anthropic, Gemini, Groq, xAI, custom HTTP, or custom WebSocket
  • LLM Model — Specific model (e.g., gpt-4o, claude-3-sonnet)
  • System Prompt — Instructions defining the assistant’s behavior
  • Temperature — Creativity level (0.0-1.0)

​
Voice Configuration

  • TTS Provider — ElevenLabs, Deepgram, Cartesia, Azure, Inworld, Resemble, OpenAI, Kokoro, Uplift, Murf, or Soniox
  • Voice — Specific voice ID or name
  • Speed — Speaking rate adjustment
  • Stability — Voice consistency (provider-dependent)

​
Speech Recognition

  • STT Provider — Deepgram (recommended), ElevenLabs, Azure, Uplift, Speechmatics, Telnyx, or Soniox
  • Model — Nova-3, Flux, Scribe v2, Azure, Telnyx-hosted, Soniox, or other provider models
  • Language — Primary language for recognition
  • Keywords — Important terms for better recognition

​
Call Management Settings

​
Recording

  • Enable Recording — Toggle call recording on/off
  • Recording Format — MP3 or WAV

​
Interruption Handling

  • Allow Interruptions — Let callers interrupt the assistant
  • Interruption Threshold — Sensitivity (0.0-1.0)

​
Timeouts

  • Max Call Duration — Maximum call length in minutes
  • Silence Timeout — End call after N seconds of silence

​
API Configuration

For API access, you’ll need:

​
Example API Request

​
Webhook Configuration

Configure webhooks in Settings → Webhooks to receive real-time notifications:

​
Best Practices

Shorter system prompts reduce latency. Focus on essential instructions and personality traits.
For simple tasks, faster models (GPT-4o-mini) may be better than more powerful ones.
Set up backup providers for critical assistants to ensure reliability.
Use the test call feature to verify your configuration before assigning real phone numbers.

​
Next Steps

AI Configuration

Deep dive into LLM, TTS, and STT settings

STT Providers

Tune speech-to-text provider, VAD, endpointing, and silence controls

API Reference

Full API documentation