548 frameworks. Nobody is reading that raw. Filter in the right order and each filter cuts meaningfully. - Filter one: language. Cuts the field by more than half. - Filter two: maintenance pulse. Recent commits, answered issues, a release this year. Be ruthless. - Filter three: license. Non-negotiable for commercial builds, and it removes more options than people expect. - Filter four: pluggable model layer, not welded to one provider. - Filter five: local-LLM support, if privacy or cost matters. - Now, and only now, read docs. Build the same toy project in your top two. The right choice is usually obvious within a weekend. Filters first, feelings later, data throughout. Want to skip the filtering entirely? PrivateLLM deploy sets up your private LLM on AWS for $50 plus usage.