Agent Definition with Deepgram

Deploy real-time speech-to-text agents using Deepgram's industry-leading transcription API.

Features

• Real-Time Transcription: Ultra-low latency streaming STT
• High Accuracy: Best-in-class transcription quality
• Multi-Language: Support for 30+ languages
• Speaker Diarization: Identify different speakers

Prerequisites

Infrastructure

Requires Docker-enabled environment or cloud VM.

# Add lattice user to Docker group
sudo adduser lattice docker
sudo systemctl restart lattice

API Keys

Set your Deepgram API key:

export DEEPGRAM_API_KEY="your-api-key-here"

Get your API key from: https://console.deepgram.com/

Architecture

This agent definition provisions:

• Python environment with Deepgram SDK
• Pre-configured real-time transcription pipeline
• WebSocket support for streaming audio
• Sample agent code for speech recognition

Note: Edit the Terraform to customize transcription models and features.

Getting Started

1. Create a new workspace from this template
2. Configure your Deepgram API key
3. Start building speech-to-text agents

Built with Deepgram