Configuration

Use configuration to swap vendors, tune latency, and enforce safety.

Minimal Config

transports:
  provider: twilio
  settings:
    account_sid: "${TWILIO_ACCOUNT_SID}"
    auth_token: "${TWILIO_AUTH_TOKEN}"
    public_url: "${TWILIO_PUBLIC_URL}"

vendors:
  stt:
    provider: deepgram
    settings:
      api_key: "${DEEPGRAM_API_KEY}"
      model: "nova-2"
  tts:
    provider: elevenlabs
    settings:
      api_key: "${ELEVENLABS_API_KEY}"
      voice_id: "${ELEVENLABS_VOICE_ID}"
  llm:
    provider: openai
    settings:
      api_key: "${OPENAI_API_KEY}"
      model: "gpt-4o-mini"

Defaults That Matter

Key	Default	Why it matters
pipeline.backpressure	drop	Keeps latency stable under load.
turn.min_barge_in_ms	300	How quickly interruptions cancel speech.
tools.timeout_ms	6000	Prevents stuck tool calls.
context.max_history	12	Controls token growth.
privacy.redact_pii	true	Protects artifacts by default.

Quick Decision Guide

If you need	Change
Lower latency	Reduce buffers, keep backpressure=drop.
No frame loss	Set backpressure=wait + increase capacity.
Stronger barge‑in	Lower turn.min_barge_in_ms.
Safer tools	Enable confirmations and raise timeouts.

Required Fields

• transports.provider
• vendors.stt.provider
• vendors.tts.provider
• vendors.llm.provider

Notes

• All string values support ${ENV_NAME} expansion.
• Use mock providers for local testing.