[Suggestion] OpenAI Agents SDK with Deepgram STT and TTS as tool providers (Python)

[deepgram-robot]

What to build

A working example integrating Deepgram's STT and TTS APIs as tool providers within the OpenAI Agents SDK. The agent should be able to transcribe audio files/streams and generate speech from text as part of its tool-calling workflow — for example, an agent that listens to a voice message, transcribes it, reasons about the content, and responds with synthesized speech.

Why this matters

The OpenAI Agents SDK is becoming a standard for building tool-using AI agents, and developers need voice capabilities in their agent pipelines. There's no existing example showing Deepgram as a tool provider in this framework. By providing a reference implementation, developers building OpenAI Agents SDK projects can add audio transcription and speech synthesis without evaluating alternatives — Deepgram becomes the default voice layer.

Suggested scope

• Language: Python
• Framework: OpenAI Agents SDK (latest)
• Deepgram APIs: Pre-recorded STT (Nova-3), TTS (Aura), optionally streaming STT
• Tools: transcribe_audio tool (file/URL → transcript), speak_text tool (text → audio file)
• Demo scenario: Voice message assistant that transcribes incoming audio, extracts key information, and responds with spoken audio
• Complexity: Intermediate — demonstrates tool definition, registration, and agent orchestration

Acceptance criteria

• Runnable with minimal setup (clone, add API keys, run)
• README explains the pattern clearly
• Uses current Deepgram SDK version and OpenAI Agents SDK
• Tools are properly typed with Pydantic models
• Includes sample audio for testing the transcription tool
• Shows both STT and TTS integration in a single agent workflow

---

Raised by the DX intelligence system.