1. adding another hop in front of the search provider adds latency, which as an inference company I imagine you know sucks
2. I'm not convinced you actually want to route between this many search providers, for inference you generally just want the best web search available, which obviously I'd argue is Exa :) I can understand something like Apify + Exa where you need specialized/unique data, but otherwise the routing feels like extra complexity without much upside?
1. adding another hop in front of the search provider adds latency, which as an inference company I imagine you know sucks
2. I'm not convinced you actually want to route between this many search providers, for inference you generally just want the best web search available, which obviously I'd argue is Exa :) I can understand something like Apify + Exa where you need specialized/unique data, but otherwise the routing feels like extra complexity without much upside?