Resources

A Resource is one model and everything the admin does with it.

admin.add(Resource(
    Post,
    label="Post", plural="Posts", icon="file-text", group="Content",
    list=List(...), form=Form(...), detail=Detail(...),
    access=Access(view=True, add="post.add", delete=False),
    scope=Scope.tenant("team_id"),
))

Everything is optional but the model

admin.add(Resource(Post))

That is a working list, form and detail page, derived at mount from the model’s own columns. It is the difference between an admin you can point at a new model in ten seconds and one you configure before you can look at anything.

What is derived:

Label and plural	BlogPost → “Blog post” / “Blog posts”. English regulars plus the common irregulars; plural= for anything else
Slug	/admin/blog-post
Permissions	blog-post.view, .add, .change, .delete
List	identity first, then state, then time — capped at seven columns
Filters	a search box over the text columns, a chip per state column, one date range
Ordering	newest first when there is a timestamp, else the key descending
Form	every writable column, with the timestamps in a collapsed Audit group
Detail	the form’s fields, plus a window onto each child table

Every one of those is replaced by naming it, and nothing fights a declaration that exists.

Naming things

Resource(
    ClassRoom,
    label="Class",
    plural="Classes",
    # Without these the slug is `class-room`, and so is every permission.
    slug="class",
    stem="class",
)

slug is the URL segment; stem is what the four permissions are named after. They are separate because a URL and a permission name change for different reasons.

Flags versus access

Resource(Payment, creatable=False, editable=False)

A flag is a statement about the model — a row created by a job and never by hand — and it holds however the permissions are set. Access is a statement about a person. Both are applied, and the flag wins:

Resource(Post, creatable=False, access=Access(add="post.add"))
# → nobody may add, whatever permissions they hold

A queryset of its own

queryset= is the resource’s own idea of what it manages; scope= is this person’s slice of it. Both are applied, in that order, before anything reads a row.

Resource(
    Order,
    queryset=lambda ctx, rows: rows.filter(archived=False),
    scope=Scope.tenant("team_id"),
)

Skipping either is a leak. See Permissions.

crud(), for the shape most reference tables want

from warder import crud

for model in (Tag, Category, Region):
    admin.add(crud(model, "name", "slug", group="Reference"))

A convenience, not a layer: it returns an ordinary Resource you can take apart with .with_().