Deepgram STT

The DeepgramSTTService provides high-performance, low-latency speech recognition using Deepgram's WebSocket API. It's the recommended STT provider for most production real-time voice applications.

Installation

To use Deepgram, install the required dependencies:

pip install "piopiy-ai[deepgram]"

Prerequisites

• A Deepgram account and API key (Get yours here).
• Set your API key in your environment:

  export DEEPGRAM_API_KEY="your_api_key_here"

Configuration

DeepgramSTTService Parameters

Parameter	Type	Default	Description
api_key	str	Required	Your Deepgram API key.
base_url	str	""	Custom Deepgram API base URL.
sample_rate	int	None	Audio sample rate. Defaults to transport rate.
live_options	LiveOptions	None	Deepgram-specific transcription options.

Default LiveOptions

If no live_options are provided, Piopiy uses these defaults:

Option	Default
model	"nova-3-general"
language	"en"
interim_results	True
smart_format	True
punctuate	True

Usage

Basic Setup

import os
from piopiy.services.deepgram.stt import DeepgramSTTService

stt = DeepgramSTTService(
    api_key=os.getenv("DEEPGRAM_API_KEY"),
)

With Custom LiveOptions

from deepgram import LiveOptions
from piopiy.services.deepgram.stt import DeepgramSTTService

options = LiveOptions(
    model="nova-2-medical",
    language="en-US",
    smart_format=True
)

stt = DeepgramSTTService(
    api_key=os.getenv("DEEPGRAM_API_KEY"),
    live_options=options
)

Notes

• Model Selection: nova-3 is the latest and fastest model for general use cases.
• Interim Results: Enabled by default to provide the LLM with early context, reducing perceived latency.