Release Notes - v0.1.0¶

Release Date: 2026-01-19

Overview¶

Initial release of omnivoice-deepgram, an OmniVoice STT provider for Deepgram speech-to-text services.

This package adapts the official Deepgram Go SDK to the OmniVoice interfaces, enabling real-time streaming transcription for voice agents and telephony applications.

Highlights¶

OmniVoice STT provider for Deepgram with real-time streaming transcription support

What's Included¶

Streaming Transcription¶

Full implementation of stt.StreamingProvider interface
Real-time transcription via Deepgram WebSocket API
Interim results for responsive UI feedback
Final results with word-level timing information

Audio Format Support¶

mu-law (Twilio, telephony)
A-law (European telephony)
Linear PCM (general audio)
FLAC, Opus, MP3

Voice Agent Features¶

Speech start/end detection for natural turn-taking
Speaker diarization for multi-speaker scenarios
Keyword boosting for domain-specific vocabulary

Configuration¶

OmniVoice TranscriptionConfig to Deepgram options converter
Sensible defaults for telephony (8kHz, mono, nova-2 model)

Installation¶

go get github.com/agentplexus/omnivoice-deepgram

Quick Start¶

import (
    deepgramstt "github.com/agentplexus/omnivoice-deepgram/omnivoice/stt"
    "github.com/agentplexus/omnivoice/stt"
)

// Create provider
provider, err := deepgramstt.New(deepgramstt.WithAPIKey("your-api-key"))
if err != nil {
    log.Fatal(err)
}

// Start streaming
writer, events, err := provider.TranscribeStream(ctx, stt.TranscriptionConfig{
    Model:      "nova-2",
    Language:   "en-US",
    Encoding:   "mulaw",
    SampleRate: 8000,
})

// Send audio, receive transcripts
go func() {
    defer writer.Close()
    io.Copy(writer, audioSource)
}()

for event := range events {
    if event.IsFinal {
        fmt.Println(event.Transcript)
    }
}

Dependencies¶

github.com/agentplexus/omnivoice v0.2.0 - OmniVoice interfaces
github.com/deepgram/deepgram-go-sdk/v3 v3.0.0 - Official Deepgram SDK

omnivoice-examples - Complete voice agent examples combining Deepgram STT with ElevenLabs TTS