Skip to main content

Providers Overview

SmarterAvatar integrates with multiple AI service providers. This flexibility lets you choose the best services for your needs and budget.

Provider Categories​

LLM Providers​

Large Language Models power the conversational intelligence:

ProviderBest ForRAG SupportPricing
GeminiRAG, long contextNative File Search$0.50/1M tokens
OpenAIGeneral purposeVia embeddings$2.50/1M tokens
AnthropicComplex reasoningVia embeddings$3.00/1M tokens

Configure LLM →

STT Providers​

Speech-to-Text converts voice input:

ProviderBest ForReal-timePricing
WhisperAccuracyNo$0.006/min
DeepgramLow latencyYes$0.0077/min

Configure STT →

Avatar Providers​

Avatar synthesis creates the video experience:

ProviderTypeQualityPricing
HeyGenReal-time streamingHigh~$0.10/min

Configure Avatar →

Quick Comparison​

For a typical 3-minute conversation:

ComponentBudget OptionPremium Option
LLMGemini Flash ($0.01)Claude Sonnet ($0.08)
STTWhisper ($0.02)Deepgram ($0.02)
AvatarHeyGen ($0.30)HeyGen ($0.30)
Total~$0.33~$0.40

Configuration​

All providers are configured via environment variables:

# LLM Provider Selection
LLM_PROVIDER=gemini # gemini | openai | anthropic

# STT Provider Selection
STT_PROVIDER=whisper # whisper | deepgram

# Provider API Keys
GEMINI_API_KEY=AIza...
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
DEEPGRAM_API_KEY=...
LIVEAVATAR_API_KEY=...

Cost-Optimized​

Best for high-volume, budget-conscious deployments:

LLM_PROVIDER=gemini
GEMINI_MODEL=gemini-3.0-flash
STT_PROVIDER=whisper

Quality-Optimized​

Best for enterprise deployments requiring top accuracy:

LLM_PROVIDER=anthropic
ANTHROPIC_MODEL=claude-sonnet-4-20250514
STT_PROVIDER=deepgram
DEEPGRAM_MODEL=nova-2

Latency-Optimized​

Best for real-time conversation feel:

LLM_PROVIDER=gemini
GEMINI_MODEL=gemini-3.0-flash
STT_PROVIDER=deepgram
DEEPGRAM_MODEL=nova-2

Testing Providers​

Use the admin panel to test provider connectivity:

  1. Navigate to Admin Panel → Provider Testing
  2. Click Test Connection for each provider
  3. Verify all providers show green status

Or use the CLI:

pnpm verify-deployment

This checks all configured providers and reports any issues.