[Suggestion] CrewAI multi-agent audio analysis pipeline with Deepgram STT and Audio Intelligence (Python)

[deepgram-robot]

What to build

A working example showing how to build a CrewAI multi-agent system where specialized agents collaborate on audio analysis tasks — one agent transcribes audio using Deepgram STT, another extracts insights using Audio Intelligence (topics, sentiment, entities), and a third generates a structured report. Demonstrates Deepgram as the audio backbone of an agentic AI pipeline.

Why this matters

CrewAI is one of the fastest-growing multi-agent frameworks (40,000+ stars), and developers building agentic AI systems frequently need to process audio inputs — meeting recordings, customer calls, podcast episodes. Today there's no reference showing how to integrate Deepgram into a CrewAI pipeline, so developers either skip audio entirely or build ad-hoc solutions. This example makes Deepgram the default choice for audio in CrewAI workflows.

Suggested scope

• Language: Python
• Framework: CrewAI (latest stable)
• Deepgram APIs: Pre-recorded STT (Nova-3), Audio Intelligence (topics, sentiment, summarization, entity detection)
• Agents: 3 specialized CrewAI agents — Transcription Agent, Analysis Agent, Report Agent
• Input: Audio file URL or local file
• Output: Structured JSON report with transcript, topics, sentiment analysis, key entities, and executive summary
• Complexity: Intermediate — single-file main script with crew definition

Acceptance criteria

• Runnable with minimal setup (clone, add API key, run)
• README explains the pattern clearly
• Uses current SDK version and CrewAI stable release
• Demonstrates agent-to-agent data passing with Deepgram output
• Includes sample audio file or URL for testing
• Shows how to define Deepgram as a CrewAI tool

---

Raised by the DX intelligence system.