Perplexica is the open source version of perplexity. The search engine aggregator is easy to self host, but ollama can be a bit slow when self hosted. You can use an API key if you want to outsource the language model, but you want to do the searching yourself. github.com/ItzCrazyKns/Perplexica You could consider creating a compute cluster with your devices using expo. Please let me know how it goes if you do that... github.com/exo-explore/exo