Links & Location¶

A response often needs to point at another route — a Location on 201 Created, a redirect target, or the status URL on a 202 Accepted. Because the path lives on the class, jero reverse-routes to a mounted operation from the class itself, so you never hand-format these URLs.

Two response fields carry them, alongside headers / status_code:

• location: Location | None → a single Location header (RFC 9110).
• links: Sequence[Link] → one Link header, all links joined (RFC 8288).

Location on a created resource¶

create returns 201 and points Location at read_one. The method reference WidgetResource.read_one carries the class (its path) and the operation; params fills the URL slots:

from msgspec import Struct

from jero import BaseApp, JSONResponse, Location, Resource


class Widget(Struct):
    id: str


class WidgetPath(Struct):
    widget_id: str


class WidgetResource(Resource, path="/widgets"):
    async def read_one(self, path: WidgetPath) -> Widget:        # GET /widgets/{widget_id}
        return Widget(id=path.widget_id)

    async def create(self, json: Widget) -> JSONResponse[Widget]:  # POST /widgets
        return JSONResponse(
            json=json,
            status_code=201,
            location=Location.from_operation(
                WidgetResource.read_one, path=WidgetPath(widget_id=json.id)
            ),
        )


class App(BaseApp):
    async def wire(self) -> None:
        self._include_resource(WidgetResource())


app = App()

POST /widgets with {"id": "w1"} returns 201 and Location: /widgets/w1. URLs are relative by default — always RFC-valid and free of proxy-host ambiguity. See Behind a proxy below to emit absolute public URLs instead.

Links¶

Link is the same target plus a required rel (and optional title / media_type, emitted as type=). A list joins into one Link header:

from msgspec import Struct

from jero import BaseApp, JSONResponse, Link, Resource


class Widget(Struct):
    id: str


class WidgetPath(Struct):
    widget_id: str


class WidgetResource(Resource, path="/widgets"):
    async def read_one(self, path: WidgetPath) -> Widget:
        return Widget(id=path.widget_id)

    async def create(self, json: Widget) -> JSONResponse[Widget]:
        return JSONResponse(
            json=json,
            status_code=201,
            links=[
                Link.from_operation(
                    WidgetResource.read_one, rel="self", path=WidgetPath(widget_id=json.id)
                ),
                Link.from_path(
                    "/docs/widgets", rel="help", title="Docs", media_type="text/html"
                ),
                Link.from_url("https://status.example.com", rel="status"),
            ],
        )


class App(BaseApp):
    async def wire(self) -> None:
        self._include_resource(WidgetResource())


app = App()

This emits:

Link: </widgets/w1>; rel="self", </docs/widgets>; rel="help"; title="Docs"; type="text/html", <https://status.example.com>; rel="status"

Besides from_operation, there are two literal constructors: from_path(path) — a root-relative path that picks up the app's URL base just like a reversed operation (so it goes absolute under a configured proxy) — and from_url(url) — a fully-qualified URL used verbatim, never rewritten.

Validated at construction¶

from_operation(Class.operation, path=...) validates the path Struct at construction, introspected from the operation's own path annotation — so the wrong Struct fails the instant you build the link (in a handler, or a unit test that just constructs it), with no app required:

# read_one declares `path: WidgetPath` — this raises TypeError immediately:
Location.from_operation(WidgetResource.read_one, path=WrongPath(...))

(A bare method reference can't carry the path type to pyrefly statically, so this is a hard runtime check at construction — immediate, not deferred to a served request.)

Behind a proxy¶

Behind a reverse proxy you usually want absolute public URLs instead. Set JERO_BASE_URL (a static public origin) or JERO_TRUST_FORWARDED (rebuild the origin per request from X-Forwarded-*) and every reversed URL goes absolute — no code change. Details, and the trust warning that comes with forwarded headers, are on Deployment.

Circular imports: the ref escape hatch¶

from_operation needs to import the target class. When two feature modules link to each other, that import would form a cycle. For that case — and only that case — declare a string ref on the class and address it with from_ref("ref.operation", ...):

from msgspec import Struct

from jero import BaseApp, Endpoint, JSONResponse, Link, Resource


class Job(Struct):
    id: str


class JobPath(Struct):
    job_id: str


class JobResource(Resource, path="/jobs", ref="jobs"):
    async def read_one(self, path: JobPath) -> Job:
        return Job(id=path.job_id)


class JobLinkEndpoint(Endpoint, path="/job-link"):
    # Imagine this lives in a module that can't import JobResource without a cycle.
    async def get(self) -> JSONResponse[Job]:
        return JSONResponse(
            json=Job(id="job-id"),
            links=[Link.from_ref("jobs.read_one", rel="related", path=JobPath(job_id="job-id"))],
        )


class App(BaseApp):
    async def wire(self) -> None:
        self._include_resource(JobResource())
        self._include_endpoint(JobLinkEndpoint())


app = App()

The ref form trades away some safety: the string can't carry the params type to pyrefly, so its type check is deferred to resolution rather than construction, and a typo'd or unmounted ref surfaces when the response is sent rather than at startup. So prefer from_operation — reach for ref only to break a genuine import cycle (often, extracting the shared route into a module both import is cleaner still). Duplicate ref values are a startup WiringError.

See Resources & Endpoints for how the path — the thing all of this reverses against — is declared.