Enhancing LLM Transparency with the Provider Abstraction Approach

The problem with calling an LLM directly NumPath's teacher dashboard generates per-student insights — one-sentence observations like "Emma skips borrowing in 9 of 11 recent subtraction attempts" with a suggested action. The obvious implementation is to import the Anthropic SDK, call messages.create(