Module pipeline 

The read and write pipelines.

Generic over StorageAdapter rather than taking a dyn, so the storage boundary costs nothing at runtime and a future Postgres adapter drops in by type rather than by trait object.

Schema handling here is deliberately simple for 0.1.0: every call loads all schemas from the adapter. Upstream caches instead, and caching is deferred here rather than skipped by accident. When it lands, the staleness window is a decision to make deliberately and not a tuning knob: a schema read from cache decides what a write may contain and what a caller may see, so a tightened rule that has not propagated yet is still being enforced in its old, looser form. Correctness and authorization are the same question here.

Structs

CreateResponse

What a create returns to the client: {objectId, createdAt} and nothing else.

UpdateResponse

What an update returns: {updatedAt}.

Functions

count

create

Create an object.

delete

Delete one object by id.

find

Find objects.

get

Fetch one object by id.

update

Update one object by id.