Skip to main content

Provider Configuration

Rocky has two independent provider systems: one for Chat (text) and one for Voice (realtime audio). Each can be configured separately with different providers, accounts, and models.

Two Provider Systems​

  • Chat Provider — Handles text-based conversations, task planning, and tool execution
  • Voice Provider — Handles real-time voice conversations, the primary interaction mode

Only one Chat provider and one Voice provider can be active at a time, but you can configure multiple instances and switch between them.

Provider / Account / Model​

Each provider system follows a three-layer architecture:

Provider (OpenAI, Anthropic, Gemini, ...)
  └── Account (your API key)
       └── Model (GPT-4o, Claude 3.7, ...)

You can configure multiple accounts per provider and switch between models freely.

Chat Providers​

Chat providers handle all text-based interactions.

OpenAI​

Anthropic​

Azure OpenAI​

  • Models: GPT-4o (Azure deployment)
  • Setup: Requires Azure resource name, deployment name, API version, and API key

Google Gemini​

  • Models: Gemini 2.5 Pro, Gemini 2.5 Flash
  • API Key: From Google AI Studio

Groq​

  • Models: Llama 3.3 70B
  • API Key: From Groq console

xAI​

  • Models: Grok 3 Beta
  • API Key: From xAI platform

OpenRouter​

  • Models: Multi-model proxy (access many models with one key)
  • API Key: From OpenRouter

DeepSeek​

  • Models: DeepSeek Chat
  • API Key: From DeepSeek platform

Doubao (Volcengine)​

  • Models: Doubao Seed series
  • API Key: From Volcengine platform

AIProxy​

  • Models: Proxy-based access to various models
  • Setup: Requires service URL configuration

Voice Providers​

Voice providers handle real-time audio streaming for voice conversations.

OpenAI Realtime​

The most full-featured voice provider.

  • Models: GPT Realtime Mini, GPT Realtime
  • API Key: Same as OpenAI Chat API key
  • Features: Low-latency, natural voice, multi-turn conversation

Doubao Realtime​

Optimized for Chinese language voice interactions.

  • Models: Doubao E2E Voice
  • Setup: Requires Volcengine resource ID
  • Features: Excellent Chinese language support

Gemini Live​

Google's native audio model.

  • Models: Gemini 2.5 Flash Native Audio
  • API Key: From Google AI Studio
  • Features: Native audio processing

Configuration​

  1. Open Rocky app
  2. Go to Settings > Providers
  3. Switch between the Chat and Voice tabs
  4. Tap Add Provider and select your provider type
  5. Enter your API key and configure the endpoint if needed
  6. Tap to activate the provider you want to use
tip

You can mix and match — for example, use Anthropic Claude for Chat and OpenAI Realtime for Voice. Each system is completely independent.