isUsableModel

isUsableModel(llm): boolean

Defined in: packages/core/src/config/loader.ts:1450

CFG-75 — is this llm config value ALREADY usable as a model, with nothing left for the provider layer to build?

Decided by what the value can DO — answer a call — and never by what it looks like. A model a user’s own configure() built keeps its methods on its prototype and has no own type property, so a type-key test calls exactly the documented shape unusable; that is the assumption this predicate exists to replace. It is the positive half of needsProviderRouting, kept as one definition rather than two so the two questions cannot drift apart.

The two references to needsProviderRouting here are code spans and not @links on purpose: this predicate is exported and that one is not, so a link would render as a broken reference on every page that documents this function. Exporting a second symbol to satisfy a cross-reference would widen the public surface to fix a docs warning, which is the wrong way round.

It is NOT the negation of needsProviderRouting: that is false for an ABSENT llm too, so a branch that accepted !needsProviderRouting(llm) would accept a config with no model at all. Ask this one when the question is “may this be used?”, and ask it FIRST wherever a type test could also match — an object that answers calls and happens to carry a type field is a working model the user built, and sending it to the provider layer would quietly swap it for another.

CFG-61 — exported, because the loader is not the only place that has to know whether a model actually resolved. The AG-UI server’s /health and /info answer that same question about the config they were handed, and a second copy of the test is how the two would come to disagree: the loader would route a value the server called a model, or the server would report ok about a value the loader would have sent away to be built. One definition, asked from both sides.

Parameters

llm

unknown

Returns

boolean