Providers

Providers are pluggable adapters for STT, TTS, LLM, and Transport.

Selection Checklist

• Latency: streaming response time.
• Language coverage: target locales.
• Reliability: retries + uptime.
• Cost: per minute / per token.

STT

Deepgram

Settings (from examples/hvac/main.go):

• api_key, model, language, sample_rate, encoding, interim, vad_events.

Mock STT

Use for local testing without credentials.

TTS

ElevenLabs

Settings:

• api_key, voice_id, model_id, output_format, sample_rate.

Mock TTS

Use for deterministic tests.

LLM

OpenAI

Settings:

• api_key, model, base_url, use_circuit_breaker, circuit_threshold, circuit_cooldown_ms.

Mock LLM

Use for deterministic responses.

Transport

Twilio

Settings:

• account_sid, auth_token, public_url, voice_path, ws_path, status_callback_path.

Mock Transport

In‑memory transport for tests.

Add a Custom Provider

1. Implement the adapter interface in pkg/adapters/stt, pkg/adapters/tts, or pkg/llm.
2. Build a factory from ranya.Config.
3. Register it in ProviderRegistry before NewEngine.