## Outcome Make provider selection explicit, stable, permissioned, and testable through typed Neuron identities and bindings. ## Acceptance criteria - [ ] Represent each model and major version as a distinct identity. - [ ] Prevent silent substitution between identities. - [ ] Return typed unavailability and failure outcomes. - [ ] Support named account bindings. - [ ] Require permission checks for rebinding. - [ ] Provide deterministic test doubles. ## Non-goals - Automatically substituting models, versions, or account bindings.
Outcome
Make provider selection explicit, stable, permissioned, and testable through typed Neuron identities and bindings.
Acceptance criteria
Non-goals