I'm happy to help you implement Retrieval-Augmented Generation (RAG) functionality in a real-time API, similar to RAGLookup in STT/LLM/TTS pipelines. Here are some approaches you might find useful: • Using `on_user_turn_completed`: This lifecycle hook triggers right after the user finishes speaking but before the agent responds. It’s a great spot to inject RAG-retrieved context directly into the conversation, perfect for keeping things low-latency. • Using Function Tools: You can define a function tool that performs RAG retrieval only when needed. This is especially handy if not every user query requires a RAG lookup or if you want to add some logic before fetching data. • Using

llm_node()

for Manual Context Injection: This gives you full control over how you format prompts and manage context, which is fantastic for more complex RAG workflows. For some practical examples, I recommend checking out these resources: • LiveKit Docs RAG Example: Shows low-latency lookups using

on_user_turn_completed

. • LlamaIndex RAG Example: Demonstrates integration with LlamaIndex for advanced RAG workflows. If you have any questions or need a hand with the implementation, please don’t hesitate to reach out—I’m here to support you every step of the way! Sources: Pipeline nodes and hooks | LiveKit Docs | External data and RAG | LiveKit Docs | rag/README.md | examples/voice_agents/llamaindex-rag/retrieval.py | delay-handling/README.mdoc

It seems like the

on_user_turn_completed

callback might not be detected in your real-time API agent for a few common reasons. Let’s walk through them together: • Turn Detection Configuration: It’s important to make sure turn detection is properly set up in your agent, such as using VAD or LiveKit’s turn detector plugin. If the realtime model handles turn detection on the server side, the callback won’t be triggered. • Server-Side Turn Detection: When server-side turn detection is enabled in the realtime model, the callback won’t run. For

on_user_turn_completed

to work, turn detection needs to happen within the agent itself. • Speech Interruption Settings: If your agent is generating speech and interruptions aren’t allowed, the callback might be skipped. • Speech Scheduling Paused: Also, if speech scheduling is paused, the callback won’t execute. Could you please double-check your agent’s turn detection setup to ensure it’s not relying on server-side detection? If you’d like, I’m here to help you review these settings or troubleshoot further—just let me know!