[Suggestion] Amazon Bedrock voice agent with Deepgram STT/TTS and Claude LLM (Python)

[deepgram-robot]

What to build

A voice agent example using Deepgram's Voice Agent API with Amazon Bedrock as the LLM provider (Claude), demonstrating Deepgram's composable architecture within the AWS ecosystem.

Why this matters

AWS is the dominant cloud platform for enterprise voice applications, and Amazon Bedrock gives developers access to Claude, Llama, and other LLMs without managing infrastructure. Deepgram's Voice Agent API already supports Bedrock as an LLM provider, but there is no working example showing the full integration. Developers building voice agents on AWS need a reference that demonstrates how to combine Deepgram's STT accuracy and TTS quality with Bedrock's managed LLM inference — all within their existing AWS infrastructure and IAM security model.

Suggested scope

• Language: Python
• Deepgram APIs: Voice Agent API with Bedrock LLM provider, Nova-3 STT, Aura/Flux TTS
• AWS services: Amazon Bedrock (Claude as LLM), IAM credentials
• Architecture: Deepgram Voice Agent WebSocket session configured with aws_bedrock think provider; function calling for external tool use; microphone input via PyAudio or sounddevice
• Features: Voice agent conversation loop, function calling demo (e.g., weather lookup), Bedrock model selection, proper AWS credential handling
• Complexity: Medium — requires AWS account with Bedrock access + Deepgram API key

Acceptance criteria

• Runnable with minimal setup (clone, add API key + AWS credentials, run)
• README explains the pattern clearly
• Uses current SDK version
• Demonstrates Deepgram's composable architecture (swap LLM provider without changing STT/TTS)
• Includes at least one function calling example
• Handles AWS credential configuration via standard boto3 patterns (env vars, profiles, IAM roles)

---

Raised by the DX intelligence system.