What’s the best approach for integrating a chatbot into an existing app?

I’m working on an app and want to integrate a chatbot that understands users’ natural language requests and provides relevant, personalized responses based on existing data.

I already have a MongoDB database containing scraped content from various online sources.

What would be the best approach to build this?

RAG with embeddings and a vector database?

Querying MongoDB directly using filters?

An LLM with tool/function calling?

Or a hybrid approach?

I’m interested in how you’d design the architecture, retrieve relevant information, and generate accurate responses grounded in the available data.

What technologies and architecture would you recommend, and why? I’d appreciate hearing about your experience building similar systems.