Promises are one of those features misunderstood for 2 reasons.
Several leading initial library implementations were broken from day 1, and couldn't/wouldn't later introduce breaking changes to existing users.
Another factor dragging down the impression of Promises: Google search seems to love returning 5+ year old articles. They almost universally suck.
Even the stuff I wrote back then was awful.
New articles are not necessarily spared, they often suffer from a number of issues: Overly Simplistic examples, fragile function boundaries, leaky variables, arbitrary 'design.'
Switching LLM models, providers and options should be easy.
> Was it max_tokens? Or max_output_tokens? top_p? topP? Or top_k?
disc: author share.