@rangojs/router 0.0.0-experimental.349e5728 → 0.0.0-experimental.358c376d

package/skills/loader/SKILL.md

@@ -1,7 +1,7 @@
 ---
 name: loader
-description: Define data loaders for fetching data in routes with createLoader
-argument-hint: [loader]
+description: Define data loaders for fetching data in routes with createLoader. Use when pages need per-request data that stays fresh, data should stream while the page renders, client components need reactive server data, a loader should throw notFound()/redirect(), set page meta/breadcrumbs from loader data (handle writes), or loader data must be guaranteed in the SSR'd document (stream:"navigation").
+argument-hint: "[loader]"
 ---
 
 # Data Loaders with loader()
@@ -11,6 +11,13 @@ Loaders fetch data on the server and stream it to the client. For mutations
 `/server-actions`. Loaders re-resolve after an action runs, so the typical
 flow is _action mutates → loader re-reads → UI updates_.
 
+## Not this skill if…
+
+- You want to mutate state — mutations are `"use server"` actions: see
+  `/server-actions`. Loaders read per-request live data.
+- You want to cache a function's return value — loaders are fresh every request
+  by default; caching one function is `"use cache"`: see `/use-cache`.
+
 ## Creating a Loader
 
 ```typescript
@@ -91,6 +98,20 @@ path("/product/:slug", ProductPage, { name: "product" }, () => [
 ]);
 ```
 
+> **Client refresh `key` vs. server `cache({ key })` vs. `revalidate()`.** Three
+> different "what refreshes" knobs that are easy to confuse:
+>
+> - `useLoader(Loader, { key })` / `useFetchLoader(Loader, { key })` — a
+>   **client** refresh identity. It groups which mounted reads of one loader
+>   refresh together when one calls `load()`. It never touches the server
+>   request. For refreshing **different** loaders together, tag them with
+>   `{ refreshGroup }` (one name or several) and call `useRefreshLoaders()(name)`
+>   (plain GET only). See the hooks skill ("Scoping refetch with a `key`" and
+>   "Refreshing multiple loaders together").
+> - `cache({ key })` — a **server** cache identity (storage hit/miss/ttl/swr).
+> - `revalidate()` — which **server** segments/loaders recompute during
+>   navigation and action refreshes.
+
 DSL loaders are the **live data layer** — they resolve fresh on every
 request, even when the route is inside a `cache()` boundary. The router
 excludes them from the segment cache at storage time and re-resolves them
@@ -126,6 +147,11 @@ same memoized result — loaders never run twice per request.
 - The handler output depends on the loader data. If the route is inside
   `cache()`, the handler is cached with the loader result baked in —
   defeating the live data guarantee.
+- The same holds under a PPR shell capture (`/ppr`): handler consumption is
+  the BAKED lane — the loader executes at capture (identity reads permitted)
+  and the rendered value is a capture-time copy; `useLoader` client-side is
+  the live lane. One rule across `cache()`, `"use cache"`, and PPR: the
+  consumption-lane rule (`/rango` → Invariants).
 - Non-cacheable variable reads (`createVar({ cache: false })`) inside the
   handler still throw, even if the data came from a loader.
 - Prefer DSL `loader()` + client `useLoader()` for data that depends on
@@ -146,23 +172,23 @@ Loaders receive the same context shape as route handlers.
 
 ### Full field surface
 
-| Field          | Type                           | Notes                                                                                               |
-| -------------- | ------------------------------ | --------------------------------------------------------------------------------------------------- |
-| `params`       | `TParams`                      | Merged route + explicit loader params; overridable by fetchable `load({ params })`.                 |
-| `routeParams`  | `Record<string, string>`       | Server-trusted route params from URL pattern matching; cannot be overridden.                        |
-| `request`      | `Request`                      | The incoming `Request` (headers, method, body, `signal` for abort).                                 |
-| `url`          | `URL`                          | Parsed request URL.                                                                                 |
-| `pathname`     | `string`                       | URL pathname (shortcut for `ctx.url.pathname`).                                                     |
-| `searchParams` | `URLSearchParams`              | Shortcut for `ctx.url.searchParams`.                                                                |
-| `search`       | `ResolveSearchSchema<TSearch>` | Typed query params when a search schema is declared on the route; `{}` otherwise.                   |
-| `env`          | `TEnv`                         | Plain bindings from `createRouter<TEnv>()` (DB, KV, secrets, etc.).                                 |
-| `get`          | `(key \| ContextVar) => value` | Reads variables/context-vars set by middleware.                                                     |
-| `use`          | `(loader \| handle) => T`      | Access another loader's data (Promise) or a handle's collected data (after `await ctx.rendered()`). |
-| `rendered`     | `() => Promise<void>`          | **Experimental.** DSL loaders only — waits for non-loader segments before reading handle data.      |
-| `method`       | `string`                       | HTTP method. `"GET"` for SSR loader runs; reflects real method for fetchable loaders.               |
-| `body`         | `TBody \| undefined`           | Parsed request body for fetchable POST/PUT/PATCH/DELETE calls.                                      |
-| `formData`     | `FormData \| undefined`        | Present when a fetchable loader is invoked via form submission.                                     |
-| `reverse`      | `ScopedReverseFunction`        | Generate type-checked URLs from route names (same scoped semantics as route handlers).              |
+| Field          | Type                           | Notes                                                                                                                                                   |
+| -------------- | ------------------------------ | ------------------------------------------------------------------------------------------------------------------------------------------------------- |
+| `params`       | `TParams`                      | Merged route + explicit loader params; overridable by fetchable `load({ params })`.                                                                     |
+| `routeParams`  | `Record<string, string>`       | Server-trusted route params from URL pattern matching; cannot be overridden.                                                                            |
+| `request`      | `Request`                      | The incoming `Request` (headers, method, body, `signal` for abort).                                                                                     |
+| `url`          | `URL`                          | Parsed request URL.                                                                                                                                     |
+| `pathname`     | `string`                       | URL pathname (shortcut for `ctx.url.pathname`).                                                                                                         |
+| `searchParams` | `URLSearchParams`              | Shortcut for `ctx.url.searchParams`.                                                                                                                    |
+| `search`       | `ResolveSearchSchema<TSearch>` | Typed query params when a search schema is declared on the route; `{}` otherwise.                                                                       |
+| `env`          | `TEnv`                         | Plain bindings from `createRouter<TEnv>()` (DB, KV, secrets, etc.).                                                                                     |
+| `get`          | `(key \| ContextVar \| handle)` | Reads middleware variables/context-vars — or READS a handle's collected data, after `await ctx.rendered()`.                                             |
+| `use`          | `(loader \| handle) => T`      | Access another loader's data (Promise), or WRITE a handle: `ctx.use(Meta)({ title })` returns the push function — handler parity. Reads moved to `get`. |
+| `rendered`     | `() => Promise<void>`          | **Experimental.** DSL loaders only — waits for all non-loader segments (including `loading()` streaming handlers) to settle before reading handle data. Not with `stream: "navigation"` (cycle; throws). |
+| `method`       | `string`                       | HTTP method. `"GET"` for SSR loader runs; reflects real method for fetchable loaders.                                                                   |
+| `body`         | `TBody \| undefined`           | Parsed request body for fetchable POST/PUT/PATCH/DELETE calls.                                                                                          |
+| `formData`     | `FormData \| undefined`        | Present when a fetchable loader is invoked via form submission.                                                                                         |
+| `reverse`      | `ScopedReverseFunction`        | Generate type-checked URLs from route names (same scoped semantics as route handlers).                                                                  |
 
 ### Example
 
@@ -185,7 +211,7 @@ export const ProductLoader = createLoader(async (ctx) => {
   // Request headers
   const auth = ctx.request.headers.get("Authorization");
 
-  // Variables set by middleware (from RSCRouter.Vars augmentation)
+  // Variables set by middleware (from Rango.Vars augmentation)
   const user = ctx.get("user");
 
   // Type-checked URLs for payloads. `.name` resolves within the current
@@ -235,6 +261,8 @@ export const OrderLoader = createLoader(async (ctx) => {
 Add caching or revalidation to specific loaders:
 
 ```typescript
+import * as CartActions from "./actions/cart";
+
 path("/product/:slug", ProductPage, { name: "product" }, () => [
   // Cached loader
   loader(ProductLoader, () => [cache({ ttl: 300 })]),
@@ -244,15 +272,23 @@ path("/product/:slug", ProductPage, { name: "product" }, () => [
     revalidate(() => false), // Never revalidate
   ]),
 
-  // Loader that revalidates after cart actions
+  // Loader that revalidates after cart actions (defer otherwise — keeps the
+  // permissive loader defaults for navigation and other actions intact)
   loader(CartLoader, () => [
-    revalidate(({ actionId }) => actionId?.includes("Cart") ?? false),
+    revalidate((ctx) => ctx.isAction(CartActions) || undefined),
   ]),
 ]);
 ```
 
 ### `revalidate()` return shapes
 
+> **Scope: `revalidate()` is a partial-render concern, not a cache concern.**
+> It decides whether a segment (here, a loader) re-runs and streams to the
+> client on a navigation or action — never whether a cached value is stale. The
+> cache decides hit/miss/ttl/swr independently and never reads `revalidate()`.
+> Caching a loader is a separate, opt-in step (`loader(Fn, () => [cache({...})])`).
+> See `/cache-guide` → "Two axes" and `/rango` → "The shape of rango".
+
 A `revalidate(fn)` callback can return one of four shapes. The chain
 processes revalidators in order; each call's return controls how the
 chain continues:
@@ -282,6 +318,58 @@ revalidate(() => null); // explicit defer
 If every revalidator on a segment defers, the segment-type default
 (e.g. params-changed for routes, `false` for parallels) is used.
 
+#### `|| undefined` (defer) vs `?? false` (hard) — pick deliberately
+
+A boolean return — including `false` — is a **hard** decision: it short-circuits
+the chain and overrides the segment default. `undefined` **defers** to the
+running suggestion / segment default. They are not interchangeable:
+
+```typescript
+// Defer: "revalidate on match, otherwise let the default/downstream decide."
+revalidate(({ actionId }) => actionId?.includes("Cart") || undefined);
+
+// Hard: "revalidate ONLY on match, suppress everything else."
+revalidate(({ actionId }) => actionId?.includes("Cart") ?? false);
+```
+
+This matters most for loaders, whose defaults are permissive: a loader defaults
+to revalidating on **any** action (`POST`) and on **param/search changes**
+during navigation. So `?? false` on a loader silently suppresses both — the
+loader will not refetch when you navigate to a different `:id`. Use
+`|| undefined` when you want to _add_ a revalidation signal on top of the
+sensible defaults, and reserve `?? false` for the rare case where you genuinely
+want the loader to refetch on nothing but your matched action.
+
+When **composing multiple revalidators** on one segment (see below), defer is
+mandatory: the first hard `?? false` ends the chain and the later contracts
+never run.
+
+#### Matching actions: `ctx.isAction()`
+
+To revalidate after specific server actions, match them by **reference** with
+`ctx.isAction()` rather than hand-written `actionId` substrings. A rename or
+moved file then becomes a type error instead of silently failing to match:
+
+```typescript
+import { addToCart, removeFromCart } from "../actions/cart";
+import * as CartActions from "../actions/cart";
+
+loader(CartLoader, () => [
+  revalidate((ctx) => ctx.isAction(addToCart) || undefined), // one action
+]);
+revalidate((ctx) => ctx.isAction(addToCart, removeFromCart) || undefined); // several
+revalidate((ctx) => ctx.isAction(CartActions) || undefined); // any action in the module
+```
+
+`isAction()` is a method on the revalidate predicate's **context argument** —
+there is no standalone `isAction` import; you always reach it through the callback
+parameter (`revalidate((ctx) => ctx.isAction(...))`). It returns a raw boolean, so
+pair it with `|| undefined` for the usual "revalidate on match, else defer"
+intent. It returns `false` on plain navigation and on non-matches, and resolves
+the reference the same way the router derives `actionId` (`$id` in production,
+`$$id` in dev), so it matches in both modes. The raw `actionId` string stays
+available on the same context as an escape hatch.
+
 ### Revalidation Contracts for Loader Dependencies
 
 If a loader reads `ctx.get()` data produced by an outer handler/layout, share
@@ -289,8 +377,12 @@ the same named revalidation contract across producer and consumer segments.
 
 ```typescript
 // revalidation-contracts.ts
-export const revalidateAccountScope = ({ actionId }) =>
-  actionId?.includes("src/actions/account.ts#") ?? false;
+import * as AccountActions from "./actions/account";
+
+// Match by reference with ctx.isAction() (rename-safe), and defer (|| undefined)
+// so these contracts compose — a hard `false` would short-circuit the rest.
+export const revalidateAccountScope = (ctx) =>
+  ctx.isAction(AccountActions) || undefined;
 
 layout(AccountLayout, () => [
   revalidate(revalidateAccountScope), // producer reruns
@@ -333,6 +425,67 @@ follows the same rule: at build time, loaders are skipped entirely (there is no
 real request context), and at runtime the worker resolves them fresh against
 the live database.
 
+### Parallel and streaming — latency overlaps first paint
+
+Loaders do not block the page. As the render pass begins — the pass that route
+middleware wraps, so loaders run right after middleware, not in a later
+phase — every matched loader is kicked off **concurrently** (their promises start in the
+same tick), and each result is **streamed** to the client as its own RSC Flight
+chunk rather than awaited up front. Pair a loader with `loading()` (or a
+client `<Suspense>`) and the shell paints immediately while the data streams in.
+
+This is why **"cached UI still pays full data latency" is the wrong intuition**:
+on a `cache()` hit the UI segments stream instantly from cache while the live
+loaders resolve fresh **in parallel** — data latency _overlaps_ first paint
+instead of being added on top of it. (Without a `loading()` / `<Suspense>`
+boundary a parallel loader blocks its parent, so add one to keep the overlap.)
+
+If you come from a framework where the loader is a blocking step that runs
+before the response is built, this is the shift to internalize: here the
+response starts streaming first and loader data fills in. (The one deliberate
+exception is per-loader: `loader(Def, { stream: "navigation" })` awaits that
+loader before first flush on document renders — see "`stream: "navigation"`"
+below.)
+
+### See it: `debugPerformance`
+
+Turn on the per-request performance timeline early — it is the fastest way to
+confirm loaders overlap rather than serialize, and to find the real bottleneck
+locally instead of guessing:
+
+```typescript
+const router = createRouter({ document: Document, debugPerformance: true });
+```
+
+Or enable it per-request from middleware (e.g. only when `?debug` is present) by
+calling `ctx.debugPerformance()` **before** `await next()`. Each HTML request
+then prints a shared-axis waterfall (and emits a `Server-Timing` header):
+
+```
+[RSC Perf] GET /product/widget (24.53ms)
+start      dur  span                          timeline
+ 0.08ms  3.20ms  route-matching               |#####...................................|
+ 3.40ms  8.70ms  ssr-render-html              |.....##############.....................|
+ 3.42ms 11.90ms  loader:…#ProductLoader       |.....###################................|
+ 3.45ms 11.40ms  loader:…#ReviewsLoader        |.....##################.................|
+ 0.00ms 24.53ms  handler:total                |########################################|
+```
+
+How to read it:
+
+- **Humans:** scan the `#` bars on the shared axis. Bars that start at the same
+  offset and run side by side are executing **in parallel** — loaders should
+  overlap `ssr-render-html` / `render:total`, not sit alone to the right of
+  everything. A lone `loader:*` bar past the render bar is serialized latency to
+  chase. `handler:total` is the whole request; `render:total` is the render pass.
+- **LLMs / programmatic:** read each row as `{ start, dur, label }`. A loader
+  overlaps paint when its `[start, start+dur]` interval intersects
+  `render:total` / `ssr-render-html`. Flag a regression when a `loader:*`
+  interval is **disjoint from and starts after** `render:total`, or when its
+  `dur` approaches `handler:total` — that loader is on the critical path instead
+  of overlapping it. Two `loader:*` rows with near-equal `start` confirm
+  parallel execution.
+
 ### Opting a Loader into Caching
 
 To cache a specific loader's data, attach a `cache()` child:
@@ -423,6 +576,8 @@ entirely (no read, no write).
 ### Per-Loader Store Override
 
 ```typescript
+import { MemorySegmentCacheStore } from "@rangojs/router/cache";
+
 const hotStore = new MemorySegmentCacheStore({ defaults: { ttl: 10 } });
 
 loader(PricingLoader, () => [
@@ -508,6 +663,115 @@ function ProductPage() {
 }
 ```
 
+## Loader Authority: notFound() and redirect()
+
+A loader may **throw** `notFound()` and `redirect()` — data-dependent
+authority lives with the data, so every consumer of the loader inherits the
+signal instead of re-checking existence at each read site:
+
+```typescript
+import { createLoader, notFound, redirect } from "@rangojs/router";
+
+export const ProductLoader = createLoader(async (ctx) => {
+  "use server";
+  const moved = LEGACY_SLUGS[ctx.params.slug];
+  if (moved) throw redirect(`/shop/product/${moved}`);
+
+  // Existence check BEFORE the expensive fetch: a near-instant rejection
+  // usually wins the race to first flush (see the semantics below).
+  if (!(await exists(ctx.params.slug))) notFound(`No "${ctx.params.slug}"`);
+
+  return getProduct(ctx.params.slug);
+});
+```
+
+Semantics by lane:
+
+| Signal       | Document load                                                                                                    | Client navigation                                     |
+| ------------ | ---------------------------------------------------------------------------------------------------------------- | ----------------------------------------------------- |
+| `notFound()` | Not-found UI resolves server-side (nearest `notFoundBoundary` → router option → default) and rides the envelope; the 404 STATUS is **opportunistic** — real only if the rejection beats Response construction. `stream: "navigation"` (below) makes it deterministic. | 404 UI swaps in, URL preserved, payload stays 200     |
+| `redirect()` | 200 document, then a client-side replace to the target — **no document-lane 302 from loaders**; pre-stream redirect authority belongs to middleware | Redirect envelope navigates to the target             |
+
+Session/auth gates belong in middleware (they are request-shaped, not
+data-shaped, and middleware CAN emit a real pre-stream 302). Data-dependent
+"this slug moved / does not exist" belongs in the loader.
+
+## Writing Handles from Loaders (meta, breadcrumbs)
+
+Loader bodies can WRITE handles with handler parity — `ctx.use(Handle)`
+returns the push function, legal for the whole body, streaming loaders
+included. This is how data-derived page titles and breadcrumb trails live
+where the data lives:
+
+```typescript
+import { Meta, Breadcrumbs } from "./handles";
+
+export const ProductLoader = createLoader(async (ctx) => {
+  "use server";
+  const product = await getProduct(ctx.params.slug);
+
+  ctx.use(Meta)({ title: `${product.name} — Shop` });
+  const pushCrumb = ctx.use(Breadcrumbs);
+  pushCrumb({ label: "Shop", href: "/shop" });
+  pushCrumb({ label: product.name, href: `/shop/product/${product.slug}` });
+
+  return product;
+});
+```
+
+Delivery is async **by the race model**: pushes that settle before the handler
+barrier ride the SSR handle snapshot (in the SSR'd document — `<MetaTags />`,
+`useHandle` reads); later pushes stream to the client and apply post-hydration
+on document loads (`metadata.handlesLate`) or progressively on navigations.
+A push before your slow fetch usually beats the barrier; a push derived from
+the fetched data usually does not. When it MUST be in the document, use
+`stream: "navigation"` below.
+
+Reads are the other direction and gated: `ctx.get(handle)` throws unless the
+loader first does `await ctx.rendered()` (DSL-registered loaders only —
+handler-invoked loaders cannot use `rendered()`, and a handler already
+awaiting the loader via `ctx.use()` makes it a detected deadlock).
+
+## `stream: "navigation"` — Guarantee a Loader in the Document
+
+Streaming means nothing a slow loader produces is *guaranteed* in the SSR'd
+HTML: its section SSRs as the fallback, a late handle push applies
+post-hydration, a late `notFound()` loses the status race. When the loader
+feeds something that must exist in the document — `<head>` meta via a handle,
+or a real 404 status — pass delivery options between the definition and the
+use callback:
+
+```typescript
+path("/product/:slug", ProductPage, { name: "product" }, () => [
+  loader(ProductLoader, { stream: "navigation" }, () => [cache({ ttl: 60 })]),
+  loader(RelatedLoader),   // untouched: still streams behind its boundary
+]),
+```
+
+The name says WHERE streaming still applies, not that it is disabled:
+document renders await this loader before first flush — data is settled
+(`useLoader` reads it synchronously, no fallback paints), handle pushes beat
+the barrier snapshot, and a thrown `notFound()` deterministically precedes
+Response construction (real 404, no warm-up race). Client navigations stream
+exactly as before. Scoped per LOADER: the flagged loader awaits only itself;
+siblings keep streaming.
+
+The costs and constraints:
+
+- Every document load pays the flagged loader's latency before first byte.
+  That is the point — but keep flagged loaders fast, and flag loaders, not
+  routes.
+- A flagged loader must not `await ctx.rendered()` / `ctx.get(handle)` — the
+  document render awaits the loader before the render barrier resolves, so
+  that wait is a cycle by construction; it throws a deadlock error naming the
+  fix.
+- PPR capture renders mask loaders and skip the await — the flag does not
+  bake anything into a shell (`/ppr`).
+
+Also available in `clientUrls()` route groups (`/client-urls`), where the
+loader-heavy shape makes it most useful. `LoaderOptions` is exported from the
+package root.
+
 ## Fetchable Loaders
 
 By default, loaders only run during SSR and navigation. Pass `true` as the second
@@ -531,6 +795,16 @@ export const SearchLoader = createLoader(async (ctx) => {
 }, true); // true = fetchable
 ```
 
+> **No registration needed — and no worker-entry import.** A fetchable loader
+> does not have to be registered with `loader()` in the route DSL, and it does
+> not have to be imported by any server module. Importing it into the client
+> component that calls `useFetchLoader()` / `load()` is enough. Rango discovers
+> every `createLoader(fn, true)` at build time and registers it for the
+> `_rsc_loader` endpoint, so a loader reachable only through a client component
+> still resolves in production — on both the generated entry and a hand-written
+> worker entry (e.g. a Cloudflare `worker.rsc.tsx`). You do **not** need to
+> force-import the loader in your worker entry to make it resolve.
+
 ### Fetchable Loader with Middleware
 
 Pass an options object instead of `true` to attach per-loader middleware.
@@ -606,6 +880,13 @@ export const FileUploadLoader = createLoader(async (ctx) => {
 
 Client usage — see `/hooks useFetchLoader` for the full client-side pattern.
 
+> **Refetch sharing**: when the loader is registered on the route via
+> `loader()`, a plain `load()` call (no `params`, no `body`) broadcasts
+> the new value to every component reading the same loader id —
+> `useLoader` reads in layouts, pages, and parallel slots all converge.
+> Calls with `params` or a non-GET method stay local to the call site.
+> See `/hooks` → "Shared refetch behavior" for the full contract.
+
 ## Complete Example
 
 ```typescript
@@ -638,10 +919,12 @@ export const CartLoader = createLoader(async (ctx) => {
 });
 
 // urls.tsx — register loaders in the DSL
+import * as CartActions from "./actions/cart";
+
 export const urlpatterns = urls(({ path, layout, loader, loading, cache, revalidate }) => [
   layout(<ShopLayout />, () => [
     loader(CartLoader, () => [
-      revalidate(({ actionId }) => actionId?.includes("Cart") ?? false),
+      revalidate((ctx) => ctx.isAction(CartActions) || undefined),
     ]),
 
     path("/shop/product/:slug", ProductPage, { name: "product" }, () => [

package/skills/middleware/SKILL.md

@@ -1,6 +1,6 @@
 ---
 name: middleware
-description: Define middleware for authentication, logging, and request processing in @rangojs/router
+description: Define middleware for authentication, logging, and request processing in @rangojs/router. Use when gating routes behind auth checks, logging requests, or running shared logic before a handler runs.
 argument-hint: [middleware-name]
 ---
 
@@ -10,9 +10,6 @@ Middleware runs before/after route handlers using the onion model.
 
 ## Execution Model
 
-Canonical semantics reference:
-[docs/execution-model.md](../../docs/internal/execution-model.md)
-
 There are two levels of middleware with different execution scopes:
 
 ### Global middleware (`router.use()`)
@@ -36,15 +33,22 @@ Registered inside `urls()` callback. Wraps **rendering only** -- it does NOT wra
 
 ```
 Request flow (with action):
-  global mw -> action executes -> route mw -> layout -> handler -> loaders
+  global mw -> action executes -> route mw -> render pass
 
 Request flow (no action):
-  global mw -> route mw -> layout -> handler -> loaders
+  global mw -> route mw -> render pass
 
 Progressive enhancement (no-JS form POST):
   global mw -> action executes -> route mw -> full page re-render
 ```
 
+The **render pass** resolves handler, layouts, parallels, and loaders together —
+it is not a handler-then-loaders sequence. Handler-first ordering is guaranteed
+only between a route handler and its child/orphan layouts and parallels (so
+`ctx.set` is visible); loaders run **concurrently** and stream their results, so
+their latency overlaps rendering rather than blocking it. See `/loader` →
+"Parallel and streaming".
+
 The contract is: **route middleware wraps rendering regardless of transport** (JS-enabled RSC stream or no-JS HTML). During PE re-render, route middleware observes action-set state (cookies, context variables) the same way it does during JS-enabled post-action revalidation.
 
 Revalidation is still partial. Route middleware wraps the render pass that
@@ -56,15 +60,21 @@ data itself.
 ### Revalidation Contracts with Middleware-Backed Trees
 
 Middleware can establish request-level context (`ctx.set`) for segments that
-execute in the current render pass. It does not change partial revalidation
-boundaries between handler/layout/parallel segments.
+execute in the current render pass. Because route middleware wraps **every**
+render pass — normal renders, post-action revalidation, PE re-renders — its
+variables are never stale: middleware is the safest `ctx.set` rung on the
+data-passing ladder (`/rango` → "Passing data down the tree"). But it does
+not change partial revalidation boundaries between handler/layout/parallel
+segments.
 
 For shared segment data, use named revalidation contracts on both the producer
 and consumer segments, even when middleware is present in the chain.
 
 ```typescript
-export const revalidateCartData = ({ actionId }) =>
-  actionId?.includes("src/actions/cart.ts#") ?? false;
+import * as CartActions from "./actions/cart";
+
+export const revalidateCartData = (ctx) =>
+  ctx.isAction(CartActions) || undefined;
 
 layout(CartLayout, () => [
   middleware(cartRenderMiddleware),
@@ -187,14 +197,18 @@ export const myMiddleware: Middleware = async (ctx, next) => {
   ctx.request; // Request object
   ctx.url; // Parsed URL
   ctx.params; // Route parameters
+  ctx.build; // in middleware: true only during Prerender + ppr build-shell capture (plain Prerender does not run middleware)
 
   // Access platform bindings (plain bindings from createRouter<TEnv>())
   ctx.env.DB; // D1Database
   ctx.env.KV; // KVNamespace
 
-  // Set variables for downstream handlers (typed via RSCRouter.Vars)
+  // Set variables for downstream handlers (typed via Rango.Vars)
   ctx.set("user", { id: "123", name: "John" });
 
+  // Opt the current request out of PPR shell lookup/capture.
+  ctx.dynamic();
+
   // Continue to next middleware/handler
   await next();
 
@@ -233,8 +247,37 @@ const Dashboard: Handler<"dashboard"> = (ctx) => {
 ```
 
 This works alongside `ctx.get("key")` / `ctx.set("key", value)` (global typing
-via RSCRouter.Vars augmentation). Use `createVar` for route-local or feature-scoped
-data; use RSCRouter.Vars for app-wide middleware state.
+via Rango.Vars augmentation). Use `createVar` for route-local or feature-scoped
+data; use Rango.Vars for app-wide middleware state.
+
+## Build-Time PPR Middleware
+
+Normal `Prerender` Flight payload collection does not run middleware: there is
+no request to wrap. The exception is `Prerender` + `ppr` build-shell capture.
+After the Flight payload exists, the shell producer replays global and route
+middleware for each generated URL before it captures HTML.
+
+In that build-shell pass:
+
+- `ctx.build === true`;
+- `ctx.waitUntil()` is inert;
+- `ctx.dynamic()` skips the baked shell for that URL;
+- context variables set by middleware are visible to the shell render.
+
+Use this to keep side effects predictable:
+
+```typescript
+export const commerceMiddleware: Middleware = async (ctx, next) => {
+  if (ctx.build) {
+    ctx.dynamic(); // leave this shell to runtime PPR
+    return next();
+  }
+
+  const session = await commerce.auth(ctx.request);
+  ctx.set("session", session);
+  return next();
+};
+```
 
 ## Redirect with State in Middleware