[Suggestion] Smolagents (Hugging Face) voice-enabled agent with Deepgram STT and TTS (Python)

[deepgram-robot]

What to build

A working example of a Hugging Face Smolagents agent that can hear and speak using Deepgram STT and TTS. Smolagents is Hugging Face's lightweight agent framework that is rapidly gaining adoption in the ML community, but there is no published example showing how to add voice capabilities to a Smolagents agent.

Why this matters

ML engineers and researchers using the Hugging Face ecosystem build agents with Smolagents but lack a path to voice-enable them. Adding Deepgram STT as an input tool and TTS as an output tool would let developers create voice-interactive research assistants, data analysis agents, and conversational ML tools. This integration surfaces Deepgram in the Hugging Face ecosystem — one of the largest ML developer communities.

Suggested scope

• Language: Python
• Framework: Smolagents (huggingface/smolagents)
• Deepgram APIs: Speech-to-Text (streaming or pre-recorded), Text-to-Speech (Aura)
• What it demonstrates: A Smolagents agent with DeepgramSTTTool and DeepgramTTSTool custom tools, enabling voice input for queries and spoken output for responses
• Complexity: Medium — Python-only, CLI or simple web interface

Acceptance criteria

• Runnable with minimal setup (clone, add API key, run)
• README explains the pattern clearly
• Uses current SDK version
• Deepgram tools are implemented as proper Smolagents Tool subclasses
• Works with both local microphone and audio file input
• Demonstrates multi-step agent reasoning with voice I/O

---

Raised by the DX intelligence system.