What are you doing for a reliable, well designed local model based AI search?

Ive been using OpenWebUI's inbuilt web search and I can't put my finger on it, but it's not quite as polished, reliable, seamless as using Claude or Gemini. And its often slow beyond a limit that makes me subconsciously not want to use it - I understand its unlikely to get to cloud speed without making huge compromises in model quality/quantization, but im ok with something like a minute or two max.

So curious what people are doing for local model based AI search, specifically:

• What UI/integration?
• What search provider?
• Which model? Smallest, most aggressive quant that is still producing reliable info and good summaries even with many search results being injected to the LLM