[Suggestion] AG2 multi-agent voice orchestration pipeline (Python)

[deepgram-robot]

What to build

A working example of an AG2 (the actively maintained successor to AutoGen) multi-agent system with voice capabilities powered by Deepgram. AG2 enables complex multi-agent workflows where specialized agents collaborate — adding Deepgram voice I/O would demonstrate voice-driven agent orchestration patterns.

Why this matters

Developers building multi-agent systems with AG2 need a way to add voice interaction so that users can speak to an orchestrator agent that delegates tasks to specialist agents (e.g., a research agent, a coding agent, a data agent) and speaks back the results. This pattern is increasingly important for enterprise AI assistants, customer service automation, and interactive research tools.

Suggested scope

• Language: Python
• Framework: AG2 (ag2ai/ag2, latest)
• Deepgram APIs: Speech-to-Text (streaming), Text-to-Speech (Aura)
• What it demonstrates: A multi-agent conversation where a user speaks to an orchestrator agent, which delegates to 2-3 specialist agents, aggregates responses, and speaks the answer back. Each agent has a distinct role and the conversation flow is visible.
• Complexity: Medium-high — multi-agent setup with voice I/O

Acceptance criteria

• Runnable with minimal setup (clone, add API key, run)
• README explains the multi-agent architecture and voice flow clearly
• Uses current SDK version
• At least 3 agents with distinct roles demonstrating delegation
• Voice input triggers the orchestration pipeline
• Agent-to-agent communication is visible in logs/console output

---

Raised by the DX intelligence system.