feat: mount discovered module UI in a minimal Ekklesia browser host

Phase 6C pass 2. Every link in the module UI chain existed except the one
that closes it: a module could be authored, built, validated, published at a
browser URL, and loaded by a generic loader — and nothing in Ekklesia did any
of it. There was no browser application at all.

Two pieces were missing. A browser could not discover what to mount, and it
must not work that out for itself.

GET /api/v1/module-ui is the inventory: per publishable module, its id and
version, the server-produced artifact URL, the server-produced root of the
module's own API, and the subset of that module's declared permissions the
caller holds. Five fields, each because building a ModuleUiHostContext needs
it — no filesystem paths, and no name/icon/folder/order, because navigation
metadata is a design that does not exist yet. Its module set is the ADR 0025
publication registry rather than a rescan, so the inventory and the artifact
route cannot disagree. It requires a new Core permission, core.modules.ui.read
(baseline authenticated), because it is precisely the enumeration the anonymous
artifact route deliberately withholds.

Both URLs are server-produced. moduleUiArtifactUrl already existed;
moduleApiBaseUrl is now exported and moduleRoutePath builds on it, so each URL
shape has exactly one author. Without it the shell would hardcode
/api/v1/modules/<id> — the prefix ADR 0022 forbids a module UI from spelling,
reappearing in the host instead.

surfaces/shell (@ekklesia/shell) is the repository's first surface: a small
React application built to static files by its own esbuild script. It resolves
the principal, fetches the inventory, builds the host context entirely from
server-derived data, and hands the artifact URL to @ekklesia/module-ui-host.
One frame per selection — keyed by module identity — defeats the stale-load
race; the host owns the container, the loading state, and every failure, so a
broken module costs its own panel rather than the shell.

The shell statically imports no module, and both levels are checked: no source
import may name a module, and the emitted bundle must contain exactly one
dynamic import (the loader's) with no module id, no /api/v1/modules, no eval,
and no blob URL. Permission hints come from Core's own evaluator and stay
presentation-only — the end-to-end test has a UI holding no grant call its API
anyway and still receive 403.

Surfaces may now depend only on browser-safe shared roles; @ekklesia/contracts
is deliberately excluded, since its barrel pulls Drizzle's types and Node's
globals into the browser — the hole ADR 0024 closed one layer down.

Not the Dashboard: no navigation, grouping, icons, ordering, design system, or
routing. The selector is a list of module ids.

Co-Authored-By: Claude Opus 5 <[email protected]>

apps/api/src/app.ts

@@ -16,6 +16,7 @@ import { cors } from "hono/cors";
 import { AUTH_REQUEST_MAX_BODY_BYTES, platformBodyLimit } from "./body-limits.js";
 import { createGrantsRoutes } from "./grants-routes.js";
 import { canonicalizeVaryHeader, mountModuleHttpRoutes } from "./module-http-routes.js";
+import { createModuleUiInventoryRoutes } from "./module-ui-inventory.js";
 import { createModuleUiRoutes } from "./module-ui-routes.js";
 
 export interface CreateApiAppAuthorizationOptions {
@@ -179,6 +180,19 @@ export function createApiApp(options: CreateApiAppOptions): Hono {
   // that prefix rather than nested inside it.
   app.route("/", createModuleUiRoutes({ registry: options.moduleUi }));
 
+  // The browser host's bootstrap query, at the collection root of that same prefix. Unlike the
+  // artifact route above it is authenticated (`core.modules.ui.read`) and per-principal, because it
+  // is the module enumeration ADR 0025 deliberately withheld from the anonymous artifact route.
+  app.route(
+    "/",
+    createModuleUiInventoryRoutes({
+      evaluator: options.authorization.evaluator,
+      moduleUi: options.moduleUi,
+      registry: options.authorization.registry,
+      runtime,
+    }),
+  );
+
   mountModuleHttpRoutes(app, options.moduleRoutes, {
     evaluator: options.authorization.evaluator,
     registry: options.authorization.registry,

apps/api/src/module-http-routes.ts

@@ -45,13 +45,30 @@ export class ModuleRouteConfigurationError extends Error {
   }
 }
 
+/**
+ * The root a module's own HTTP API is mounted at — the one place that URL shape is constructed.
+ *
+ * Exported because a browser host has to resolve `context.api.request("records")`
+ * (`ModuleUiApiClient`, `@ekklesia/module-ui-sdk`) to a real Ekklesia URL, and a module UI must
+ * never build that prefix itself: it is Ekklesia's private deployment detail, and a module that
+ * hardcoded it could not be hosted anywhere else (ADR 0022). The server hands the resolved base to
+ * the host through the module UI inventory (`./module-ui-inventory.js`) rather than letting the
+ * browser re-spell this template, so this prefix can change without touching either.
+ *
+ * Percent-encoded for the same reason `moduleUiArtifactUrl` is: a value that somehow did not come
+ * from a validated manifest can only ever become one segment's worth of text, never a new segment.
+ * A real module id (lowercase kebab-case, `moduleIdSchema`) is unaffected by the encoding.
+ */
+export function moduleApiBaseUrl(moduleId: string): string {
+  return `${MODULE_ROUTE_PREFIX}/${encodeURIComponent(moduleId)}`;
+}
+
 /** "/" is the module's own root and maps to the bare module prefix, never a required trailing
  * slash; every other local path appends directly. Root and non-root therefore never collide, and
  * `/api/v1/modules/<module-id>/` (the root path with a trailing slash) is never produced here. */
 function moduleRoutePath(moduleId: ModuleId, routePath: string): string {
-  return routePath === "/"
-    ? `${MODULE_ROUTE_PREFIX}/${moduleId}`
-    : `${MODULE_ROUTE_PREFIX}/${moduleId}${routePath}`;
+  const base = moduleApiBaseUrl(moduleId);
+  return routePath === "/" ? base : `${base}${routePath}`;
 }
 
 function sanitizedModuleHandlerFailure(c: Context): Response {

apps/api/src/module-ui-inventory.ts

@@ -0,0 +1,193 @@
+import type { EkklesiaPrincipal, PermissionDeclaration } from "@ekklesia/contracts";
+import { CORE_MODULES_UI_READ } from "@ekklesia/core/authorization";
+import type { ModuleUiPublicationRegistry, PublishableModuleUi } from "@ekklesia/core/modules";
+import { Hono } from "hono";
+import type { Context } from "hono";
+import {
+  type AuthorizationMiddlewareOptions,
+  getAuthorizedPrincipal,
+  requirePermission,
+} from "./authorization-middleware.js";
+import { moduleApiBaseUrl } from "./module-http-routes.js";
+import { MODULE_UI_ASSET_PREFIX, moduleUiArtifactUrl } from "./module-ui-routes.js";
+
+/**
+ * The collection root of the module UI resource: `/api/v1/module-ui` lists what a browser host may
+ * mount, and `/api/v1/module-ui/<module-id>/<version>` (ADR 0025) serves one artifact's bytes. One
+ * prefix, one resource, two representations.
+ */
+export const MODULE_UI_INVENTORY_PATH = MODULE_UI_ASSET_PREFIX;
+
+/**
+ * Exactly what a browser host needs to mount one module's UI, and deliberately nothing else.
+ *
+ * No filesystem path, package root, package name, server/schema/migrations entrypoint, manifest
+ * version, capability, dependency, or storage bucket appears here. Those are server-side facts a
+ * browser has no use for, and several of them are precisely what ADR 0025's publication route
+ * exists to keep off the wire.
+ *
+ * There is also no `name`, `icon`, `description`, `category`, `folder`, or `order`. Those are
+ * *navigation* concerns, and ADR 0021 established that Dashboard grouping is presentation
+ * configuration an administrator owns rather than something a module declares. A field added here
+ * now would quietly become module-owned navigation metadata before that design exists; the
+ * temporary selector renders the module id instead.
+ */
+export interface HostModuleUiDescriptor {
+  readonly id: string;
+  readonly version: string;
+  /** Where the browser fetches this module's compiled UI artifact. Produced by the server's own
+   * canonical helper (`moduleUiArtifactUrl`), never re-spelled in the browser, so publication
+   * routing can change without touching a single line of host or module code. */
+  readonly ui: { readonly url: string };
+  /** The root this module's own HTTP API is mounted at, so the host can resolve the relative paths
+   * a module UI asks for (`context.api.request("records")`). Server-produced for the same reason
+   * `ui.url` is: `/api/v1/modules/<id>` is Ekklesia's private deployment detail, and neither the
+   * module nor the browser host should be the one spelling it. */
+  readonly api: { readonly baseUrl: string };
+  /**
+   * The subset of **this module's own declared permissions** that the current principal holds,
+   * evaluated server-side by Core's one authorization evaluator.
+   *
+   * Scoped to the module deliberately: a host has no business handing one module the platform's
+   * whole authorization state, and the module's manifest already declares exactly which permissions
+   * its UI can meaningfully ask about. Granted ids only — a permission the principal lacks is simply
+   * absent, so this never enumerates a module's full permission surface to someone who holds none of
+   * it.
+   *
+   * **These are presentation hints, not authority** (ADR 0022). They exist so a UI can hide a
+   * control instead of offering one that would fail. Every module route remains enforced by
+   * `requirePermission` against the composed platform registry, and that enforcement neither reads
+   * nor trusts anything in this array.
+   */
+  readonly permissions: readonly string[];
+}
+
+export interface HostModuleUiInventory {
+  readonly modules: readonly HostModuleUiDescriptor[];
+}
+
+export interface CreateModuleUiInventoryRoutesOptions extends AuthorizationMiddlewareOptions {
+  /** The same publication registry the artifact route serves from — the single accepted/publishable
+   * module set (`createModuleUiPublicationRegistry`, `@ekklesia/core/modules`). This route performs
+   * no discovery, reads no manifest, and touches no filesystem; a module absent from this registry
+   * is absent from the inventory, by construction rather than by filtering. Named `moduleUi` rather
+   * than `registry` because `AuthorizationMiddlewareOptions.registry` is already the *permission*
+   * registry, and one route needs both. */
+  readonly moduleUi: ModuleUiPublicationRegistry;
+}
+
+/**
+ * Evaluates one module's declared permissions against the caller, returning only the granted ids.
+ *
+ * Uses Core's one evaluator, so baselines, the platform-owner bypass, grant expiry, and
+ * deny-by-default all behave exactly as they do on a real request — there is no second,
+ * browser-only interpretation of what a permission means. An infrastructure failure propagates as
+ * `AuthorizationEvaluationError` rather than being folded into "not granted", because a hint that
+ * silently read "denied" during a database outage would blank the whole UI and look like a
+ * permissions problem.
+ */
+async function grantedPermissionIds(
+  evaluator: AuthorizationMiddlewareOptions["evaluator"],
+  principal: EkklesiaPrincipal,
+  declarations: readonly PermissionDeclaration[],
+): Promise<readonly string[]> {
+  const decisions = await Promise.all(
+    declarations.map(async (declaration) => ({
+      allowed: (
+        await evaluator.authorize({
+          baseline: declaration.baseline,
+          permission: declaration.id,
+          principal,
+        })
+      ).allowed,
+      id: String(declaration.id),
+    })),
+  );
+  return decisions.filter((decision) => decision.allowed).map((decision) => decision.id);
+}
+
+async function describeModule(
+  evaluator: AuthorizationMiddlewareOptions["evaluator"],
+  principal: EkklesiaPrincipal,
+  published: PublishableModuleUi,
+): Promise<HostModuleUiDescriptor> {
+  return {
+    api: { baseUrl: moduleApiBaseUrl(published.moduleId) },
+    id: published.moduleId,
+    permissions: await grantedPermissionIds(evaluator, principal, published.permissions),
+    ui: { url: moduleUiArtifactUrl(published) },
+    version: published.version,
+  };
+}
+
+/** Mirrors `requirePermission`'s own infrastructure-failure result exactly: a permission hint that
+ * could not be determined is never reported as "not granted". Carries no module id, permission id,
+ * or caught error. */
+function sanitizedInventoryFailure(c: Context): Response {
+  c.header("Cache-Control", "no-store");
+  return c.json({ code: "AUTHORIZATION_EVALUATION_FAILED" }, 503);
+}
+
+/**
+ * The browser host's one bootstrap query: which module UIs may I mount, where do I fetch each one,
+ * where does each one's API live, and what may this viewer do inside it.
+ *
+ * ## It is not a module management API
+ *
+ * It lists mountable UIs and nothing else. There is no create, update, enable, disable, install, or
+ * configure; no module that declares no UI; and no field beyond what constructing a
+ * `ModuleUiHostContext` requires. Ekklesia has no enable/disable state to report —
+ * `ModuleLifecycleStage` is vocabulary with no runtime state machine behind it — and inventing one
+ * here would be inventing a lifecycle rather than reporting one.
+ *
+ * ## The module set comes from the publication registry, never a rescan
+ *
+ * `moduleUi.list()` is the exact accepted set ADR 0025 publishes artifacts from: modules that
+ * survived discovery and declare `entrypoints.ui`, captured after the module runtime started. This
+ * route never scans a directory, reads a manifest, or reconstructs a module from `package.json`, so
+ * the inventory and the artifact route can never disagree about what exists. Order is the
+ * registry's own discovery order, which is deterministic (topologically sorted, then by id).
+ *
+ * ## Authenticated, deliberately, unlike the artifact route
+ *
+ * ADR 0025 made the artifact bytes anonymous — fetching compiled UI code confers no authority — but
+ * was careful that the route never became an enumeration oracle: unknown module, no UI, and wrong
+ * version collapse into one indistinguishable `404`, so a caller must already know both halves of an
+ * identity to learn anything. **This endpoint is that enumeration.** Serving it anonymously would
+ * hand over exactly what that route withholds, so it requires `core.modules.ui.read` (baseline
+ * `authenticated`): any signed-in user of the installation, and nobody else.
+ *
+ * Enforcement goes through the same `requirePermission` middleware every other protected route uses
+ * — no second browser-auth mechanism — which also resolves the principal exactly once and keeps a
+ * genuine denial (`403`) distinct from an infrastructure failure (`503`).
+ *
+ * The response is per-principal (its `permissions` depend on who is asking), so it is `no-store` and
+ * varies on `Origin` and `Cookie`, exactly like `/api/v1/principal`.
+ */
+export function createModuleUiInventoryRoutes(options: CreateModuleUiInventoryRoutesOptions): Hono {
+  const routes = new Hono();
+  const requireInventoryPermission = requirePermission(
+    { evaluator: options.evaluator, registry: options.registry, runtime: options.runtime },
+    CORE_MODULES_UI_READ,
+  );
+
+  routes.get(MODULE_UI_INVENTORY_PATH, requireInventoryPermission, async (c) => {
+    c.header("Cache-Control", "no-store");
+    c.header("Vary", "Origin, Cookie");
+    // Resolved and already authorized by the middleware above; never resolved a second time here.
+    // The `authenticated` baseline guarantees this is an authenticated principal.
+    const principal = getAuthorizedPrincipal(c);
+    try {
+      const modules = await Promise.all(
+        options.moduleUi
+          .list()
+          .map((published) => describeModule(options.evaluator, principal, published)),
+      );
+      return c.json({ modules } satisfies HostModuleUiInventory, 200);
+    } catch {
+      return sanitizedInventoryFailure(c);
+    }
+  });
+
+  return routes;
+}