[Suggestion] Mem0 persistent conversation memory for voice agents (Python)

[deepgram-robot]

What to build

A working example of a Deepgram voice agent with persistent conversation memory powered by Mem0. The agent should remember facts, preferences, and context from previous conversations, creating a personalized voice assistant that improves over time.

Why this matters

Voice agents today are stateless — each conversation starts fresh with no memory of previous interactions. Developers building personal assistants, customer service bots, and coaching apps need persistent memory so the agent can say "Last time you mentioned X" or recall user preferences across sessions. Mem0 (15K+ stars) is the leading memory layer for AI agents, and combining it with Deepgram voice creates a compelling personal voice assistant pattern.

Suggested scope

• Language: Python
• Framework: Mem0 (mem0ai/mem0)
• Deepgram APIs: Voice Agent API (STT + TTS), Audio Intelligence (optional — extract entities to store as memories)
• What it demonstrates: A voice agent that stores conversation memories in Mem0, retrieves relevant memories at the start of each new conversation, and uses them to personalize responses. Should show memory persistence across separate sessions.
• Complexity: Medium — requires Mem0 setup (supports in-memory for demo)

Acceptance criteria

• Runnable with minimal setup (clone, add API key, run)
• README explains the memory architecture clearly
• Uses current SDK version
• Demonstrates memory persistence across at least 2 separate conversation sessions
• Shows memory retrieval influencing agent responses
• Includes example of entity extraction from voice stored as structured memory

---

Raised by the DX intelligence system.