Release Notes: v0.13.0¶

Release Date: 2026-07-05

Summary¶

Adds Deepgram Voice Agent as a realtime provider, enabling native voice-to-voice conversations with ~100-300ms latency through the unified GetRealtimeFactory API.

Highlights¶

Deepgram Voice Agent: Third realtime provider alongside OpenAI and Gemini
Unified API: Same GetRealtimeFactory("deepgram") pattern as other providers

Added¶

Deepgram Realtime Factory¶

import "github.com/plexusone/omnivoice"

// Get Deepgram Voice Agent factory
factory, err := omnivoice.GetRealtimeFactory("deepgram")
if err != nil {
    log.Fatal(err)
}

// Create provider with configuration
provider, err := factory.Create(&gateway.RealtimeConfig{
    APIKey:       os.Getenv("DEEPGRAM_API_KEY"),
    Instructions: "You are a helpful voice assistant.",
})

Factory Discovery¶

// List all realtime factories
factories := omnivoice.ListRealtimeFactories()
// Returns: ["openai", "gemini", "deepgram"]

// Check if factory exists
if omnivoice.HasRealtimeFactory("deepgram") {
    // Deepgram is available
}

Realtime Provider Comparison¶

Provider	Latency	Registry Name	Notes
OpenAI Realtime	~100ms	`openai`	GPT-4o voice model
Gemini Live	~200ms	`gemini`	Gemini 2.0 Flash
Deepgram Voice Agent	~100-300ms	`deepgram`	Configurable LLM/TTS

Dependencies¶

Dependency	Version
omni-deepgram	v0.7.0
omnivoice-core	v0.15.0

Installation¶

go get github.com/plexusone/omnivoice@v0.13.0

Migration Guide¶

From v0.12.x¶

No breaking changes. The new deepgram realtime factory is additive.

To use Deepgram Voice Agent:

Set DEEPGRAM_API_KEY environment variable
Call GetRealtimeFactory("deepgram") or use with gateway configuration