[Suggestion] RAG voice agent with vector database knowledge retrieval during live conversation (Python)

[deepgram-robot]

What to build

A Python voice agent example that integrates retrieval-augmented generation (RAG) with Deepgram's Voice Agent API — transcribing the user's voice query, retrieving relevant context from a vector database, and responding with grounded, accurate voice answers.

Why this matters

RAG is the dominant pattern for building knowledge-grounded AI applications, but most RAG examples are text-only. Developers building voice-first knowledge bases (customer support bots, internal help desks, product information agents) need a working example showing how to combine Deepgram real-time STT with vector similarity search and voice response generation. This pattern enables voice agents that answer questions from a custom knowledge base rather than relying solely on LLM training data.

Suggested scope

• Python, using deepgram-sdk for STT and TTS
• Vector database: Pinecone, Weaviate, or ChromaDB for knowledge retrieval
• Pipeline: voice input → Deepgram STT → embedding → vector search → LLM with context → Deepgram TTS → voice output
• Example knowledge base: product documentation or FAQ corpus
• Deepgram Voice Agent API or direct STT + TTS pipeline

Acceptance criteria

• Runnable with minimal setup (clone, add API key, run)
• README explains the pattern clearly
• Uses current SDK version
• Demonstrates end-to-end voice query → knowledge retrieval → voice answer
• Includes sample knowledge base data for testing

---

Raised by the DX intelligence system.