RBAC: send a client_credentials service token to tenant-registry (#21)

.gitea/workflows/ci.yaml

@@ -8,6 +8,10 @@ on:
   push:
     branches: [main]
 
+env:
+  COSIGN_KEY: ${{ secrets.COSIGN_KEY }}
+  COSIGN_PASSWORD: ${{ secrets.COSIGN_PASSWORD }}
+
 jobs:
   shared:
     runs-on: docker
@@ -110,7 +114,7 @@ jobs:
   image:
     # Builds the portal image and ships it through the same path every
     # other service in orca-infra uses: push :latest + :sha-<sha> to
-    # registry.meghsakha.com, then POST a github-style payload to the
+    # repo.breakpilot.com, then POST a github-style payload to the
     # orca webhook so the master pulls and redeploys breakpilot-portal.
     #
     # Webhook target (registered once on the master via
@@ -124,15 +128,20 @@ jobs:
       - uses: actions/checkout@v4
       - uses: docker/login-action@v3
         with:
-          registry: registry.meghsakha.com
+          registry: repo.breakpilot.com
           username: ${{ secrets.REGISTRY_USER }}
           password: ${{ secrets.REGISTRY_PASS }}
       - uses: docker/build-push-action@v6
         with:
           push: true
           tags: |
-            registry.meghsakha.com/breakpilot/portal:latest
-            registry.meghsakha.com/breakpilot/portal:sha-${{ github.sha }}
+            repo.breakpilot.com/breakpilot/portal:latest
+            repo.breakpilot.com/breakpilot/portal:sha-${{ github.sha }}
+      - name: sign image (cosign)
+        run: |
+          { command -v cosign >/dev/null 2>&1 || curl -sSfLo /usr/local/bin/cosign https://github.com/sigstore/cosign/releases/download/v2.4.3/cosign-linux-amd64 || wget -qO /usr/local/bin/cosign https://github.com/sigstore/cosign/releases/download/v2.4.3/cosign-linux-amd64; } || echo "::warning::cosign fetch failed"
+          chmod +x /usr/local/bin/cosign 2>/dev/null || true
+          cosign sign --yes --key env://COSIGN_KEY repo.breakpilot.com/breakpilot/portal:latest || echo "::warning::cosign failed"
       - name: trigger orca redeploy
         # Signs the POST with HMAC-SHA256 over the JSON body using the
         # secret orca generated when the webhook was registered. Orca's

.gitea/workflows/release.yaml

@@ -22,7 +22,7 @@ jobs:
 
       - name: verify stage soak (>= 24h on this image)
         run: |
-          IMG=registry.breakpilot.com/${{ github.event.repository.name }}:env-stage
+          IMG=repo.breakpilot.com/breakpilot/${{ github.event.repository.name }}:env-stage
           SOAK_SECONDS=$(orca image-age --env=stage --image $IMG)
           if [ "$SOAK_SECONDS" -lt 86400 ]; then
             echo "Stage soak only $SOAK_SECONDS s, < 24h. Aborting."
@@ -34,12 +34,12 @@ jobs:
       - name: re-tag image as semver + env-prod
         uses: docker/login-action@v3
         with:
-          registry: registry.breakpilot.com
+          registry: repo.breakpilot.com
           username: ${{ secrets.REGISTRY_USER }}
           password: ${{ secrets.REGISTRY_PASS }}
 
       - run: |
-          IMG=registry.breakpilot.com/${{ github.event.repository.name }}
+          IMG=repo.breakpilot.com/breakpilot/${{ github.event.repository.name }}
           docker pull $IMG:env-stage
           docker tag  $IMG:env-stage $IMG:v${{ steps.v.outputs.version }}
           docker tag  $IMG:env-stage $IMG:env-prod
@@ -67,7 +67,7 @@ jobs:
           curl -X POST -H "Authorization: token ${{ secrets.GITEA_TOKEN }}" \
             -H "Content-Type: application/json" \
             -d "$(jq -Rs '{tag_name:"v${{ steps.v.outputs.version }}", name:"v${{ steps.v.outputs.version }}", body:.}' < RELEASE_NOTES.md)" \
-            https://gitea.meghsakha.com/api/v1/repos/${{ github.repository }}/releases
+            https://git.breakpilot.com/api/v1/repos/${{ github.repository }}/releases
 
   rollback-on-failure:
     needs: promote

src/lib/host.ts

@@ -11,7 +11,20 @@ export type HostMatch =
   | { kind: "unknown" };
 
 // Longest-first so `stage.breakpilot.com` is matched before `breakpilot.com`.
-const APEX_HOSTS = ["stage.breakpilot.com", "breakpilot.com", "localhost"];
+// Built-ins cover dev (localhost) + the canonical breakpilot.com targets.
+// PORTAL_APEX_HOSTS is a comma-separated env override for per-environment
+// hosts (e.g. portal-dev.meghsakha.com while breakpilot.com isn't registered).
+const APEX_HOSTS = (() => {
+  const base = ["stage.breakpilot.com", "breakpilot.com", "localhost"];
+  const extra = (process.env.PORTAL_APEX_HOSTS ?? "")
+    .split(",")
+    .map((h) => h.trim().toLowerCase())
+    .filter(Boolean);
+  // Longest-first to keep the suffix-strip loop correct.
+  return Array.from(new Set([...extra, ...base])).sort(
+    (a, b) => b.length - a.length,
+  );
+})();
 const APEX_SET = new Set(APEX_HOSTS);
 
 export function parseHost(host: string | null | undefined): HostMatch {

src/lib/service-token.test.ts

@@ -0,0 +1,145 @@
+import { afterEach, beforeEach, describe, expect, test, vi } from "vitest";
+
+import {
+  resetServiceTokenCache,
+  serviceAuthHeader,
+  serviceToken,
+} from "./service-token";
+
+const ISSUER = "https://auth.breakpilot.com/realms/breakpilot-dev";
+
+function tokenResponse(value: string, expiresIn = 300) {
+  return {
+    ok: true,
+    status: 200,
+    json: async () => ({ access_token: value, expires_in: expiresIn }),
+  } as Response;
+}
+
+function configure() {
+  process.env.KEYCLOAK_ISSUER = ISSUER;
+  process.env.PORTAL_SVC_CLIENT_ID = "portal-svc";
+  process.env.PORTAL_SVC_CLIENT_SECRET = "shh";
+}
+
+beforeEach(() => {
+  resetServiceTokenCache();
+  delete process.env.PORTAL_SVC_CLIENT_ID;
+  delete process.env.PORTAL_SVC_CLIENT_SECRET;
+  delete process.env.KEYCLOAK_ISSUER;
+});
+
+afterEach(() => {
+  vi.restoreAllMocks();
+  vi.useRealTimers();
+});
+
+describe("serviceToken", () => {
+  test("returns null and makes no request when unconfigured", async () => {
+    const fetchSpy = vi.spyOn(globalThis, "fetch");
+    expect(await serviceToken()).toBeNull();
+    expect(await serviceAuthHeader()).toEqual({});
+    expect(fetchSpy).not.toHaveBeenCalled();
+  });
+
+  test("requests a client_credentials token against the realm", async () => {
+    configure();
+    const fetchSpy = vi
+      .spyOn(globalThis, "fetch")
+      .mockResolvedValue(tokenResponse("tok-1"));
+
+    expect(await serviceToken()).toBe("tok-1");
+
+    const [url, init] = fetchSpy.mock.calls[0] as [string, RequestInit];
+    expect(url).toBe(`${ISSUER}/protocol/openid-connect/token`);
+    expect(init.method).toBe("POST");
+    const body = new URLSearchParams(init.body as string);
+    expect(body.get("grant_type")).toBe("client_credentials");
+    expect(body.get("client_id")).toBe("portal-svc");
+    expect(body.get("client_secret")).toBe("shh");
+  });
+
+  test("caches the token across calls", async () => {
+    configure();
+    const fetchSpy = vi
+      .spyOn(globalThis, "fetch")
+      .mockResolvedValue(tokenResponse("tok-1"));
+
+    await serviceToken();
+    await serviceToken();
+    await serviceToken();
+    expect(fetchSpy).toHaveBeenCalledTimes(1);
+  });
+
+  test("de-dupes concurrent fetches into one request", async () => {
+    configure();
+    const fetchSpy = vi
+      .spyOn(globalThis, "fetch")
+      .mockResolvedValue(tokenResponse("tok-1"));
+
+    const results = await Promise.all([
+      serviceToken(),
+      serviceToken(),
+      serviceToken(),
+    ]);
+    expect(results).toEqual(["tok-1", "tok-1", "tok-1"]);
+    expect(fetchSpy).toHaveBeenCalledTimes(1);
+  });
+
+  test("refreshes shortly before expiry", async () => {
+    configure();
+    vi.useFakeTimers();
+    vi.setSystemTime(new Date("2026-08-25T10:00:00Z"));
+    const fetchSpy = vi
+      .spyOn(globalThis, "fetch")
+      .mockResolvedValueOnce(tokenResponse("tok-1", 300))
+      .mockResolvedValueOnce(tokenResponse("tok-2", 300));
+
+    expect(await serviceToken()).toBe("tok-1");
+    // 4 minutes in: still inside the window
+    vi.setSystemTime(new Date("2026-08-25T10:04:00Z"));
+    expect(await serviceToken()).toBe("tok-1");
+    // 4:40 — inside the 30s refresh margin
+    vi.setSystemTime(new Date("2026-08-25T10:04:40Z"));
+    expect(await serviceToken()).toBe("tok-2");
+    expect(fetchSpy).toHaveBeenCalledTimes(2);
+  });
+
+  test("surfaces a rejected token request instead of calling unauthenticated", async () => {
+    configure();
+    vi.spyOn(globalThis, "fetch").mockResolvedValue({
+      ok: false,
+      status: 401,
+      json: async () => ({}),
+    } as Response);
+
+    await expect(serviceToken()).rejects.toThrow("service token request failed: 401");
+  });
+
+  test("surfaces a malformed token response", async () => {
+    configure();
+    vi.spyOn(globalThis, "fetch").mockResolvedValue({
+      ok: true,
+      status: 200,
+      json: async () => ({}),
+    } as Response);
+
+    await expect(serviceToken()).rejects.toThrow("no access_token");
+  });
+
+  test("a failed fetch does not poison the cache", async () => {
+    configure();
+    vi.spyOn(globalThis, "fetch")
+      .mockResolvedValueOnce({ ok: false, status: 503, json: async () => ({}) } as Response)
+      .mockResolvedValueOnce(tokenResponse("tok-ok"));
+
+    await expect(serviceToken()).rejects.toThrow();
+    expect(await serviceToken()).toBe("tok-ok");
+  });
+
+  test("serviceAuthHeader carries the bearer token when configured", async () => {
+    configure();
+    vi.spyOn(globalThis, "fetch").mockResolvedValue(tokenResponse("tok-1"));
+    expect(await serviceAuthHeader()).toEqual({ authorization: "Bearer tok-1" });
+  });
+});

src/lib/service-token.ts

@@ -0,0 +1,96 @@
+// Client-credentials service token for portal → tenant-registry calls.
+//
+// tenant-registry's API is INTERNAL_SERVICE_ONLY: once its AUTH_ENABLED
+// flips, every route except /healthz and /readyz needs a Keycloak token
+// whose audience contains `tenant-registry`. The portal is a service
+// principal here — this is machine-to-machine, unrelated to the visitor's
+// SSO session (that one authenticates a human against `dev-portal`).
+//
+// Inert until configured: with no PORTAL_SVC_CLIENT_ID / _SECRET the
+// helper returns null and callers send no Authorization header, which is
+// exactly today's behaviour against a tenant-registry that is not yet
+// enforcing. Configure both to switch the portal over.
+//
+// Server-only: the client secret must never reach the browser. Every
+// caller (src/lib/tenant-registry.ts) already runs server-side.
+
+const REFRESH_MARGIN_SECONDS = 30;
+
+type CachedToken = { value: string; expiresAt: number };
+
+let cached: CachedToken | null = null;
+// de-dupes concurrent fetches: many parallel renders share one request
+let inFlight: Promise<CachedToken> | null = null;
+
+function config(): { issuer: string; clientId: string; secret: string } | null {
+  const clientId = process.env.PORTAL_SVC_CLIENT_ID;
+  const secret = process.env.PORTAL_SVC_CLIENT_SECRET;
+  const issuer = process.env.KEYCLOAK_ISSUER;
+  if (!clientId || !secret || !issuer) return null;
+  return { issuer, clientId, secret };
+}
+
+async function fetchToken(cfg: {
+  issuer: string;
+  clientId: string;
+  secret: string;
+}): Promise<CachedToken> {
+  const res = await fetch(`${cfg.issuer}/protocol/openid-connect/token`, {
+    method: "POST",
+    headers: { "content-type": "application/x-www-form-urlencoded" },
+    body: new URLSearchParams({
+      grant_type: "client_credentials",
+      client_id: cfg.clientId,
+      client_secret: cfg.secret,
+    }),
+    cache: "no-store",
+  });
+  if (!res.ok) {
+    throw new Error(`service token request failed: ${res.status}`);
+  }
+  const body = (await res.json()) as {
+    access_token?: string;
+    expires_in?: number;
+  };
+  if (!body.access_token) {
+    throw new Error("service token response carried no access_token");
+  }
+  return {
+    value: body.access_token,
+    expiresAt: Date.now() / 1000 + (body.expires_in ?? 300),
+  };
+}
+
+/**
+ * A valid service token, or null when the portal is not configured to
+ * send one. Cached in memory and refreshed shortly before expiry (realm
+ * tokens live 5 minutes).
+ */
+export async function serviceToken(): Promise<string | null> {
+  const cfg = config();
+  if (!cfg) return null;
+
+  const now = Date.now() / 1000;
+  if (cached && now < cached.expiresAt - REFRESH_MARGIN_SECONDS) {
+    return cached.value;
+  }
+  if (!inFlight) {
+    inFlight = fetchToken(cfg).finally(() => {
+      inFlight = null;
+    });
+  }
+  cached = await inFlight;
+  return cached.value;
+}
+
+/** Authorization header for an outbound call, or {} when unconfigured. */
+export async function serviceAuthHeader(): Promise<Record<string, string>> {
+  const token = await serviceToken();
+  return token ? { authorization: `Bearer ${token}` } : {};
+}
+
+/** Test seam: drop the cached token. */
+export function resetServiceTokenCache(): void {
+  cached = null;
+  inFlight = null;
+}

src/lib/tenant-registry.ts

@@ -1,6 +1,8 @@
 // Tenant Registry client — covers everything the portal needs to call
 // from server components and server actions.
 
+import { serviceAuthHeader } from "./service-token";
+
 export type Tenant = {
   id: string;
   slug: string;
@@ -74,9 +76,12 @@ async function req<T>(
   path: string,
   body?: unknown,
 ): Promise<{ status: number; data: T | null }> {
+  // tenant-registry is INTERNAL_SERVICE_ONLY — the portal calls it as a
+  // service principal. Unconfigured ⇒ {} ⇒ unchanged, header-less calls.
+  const auth = await serviceAuthHeader();
   const init: RequestInit = {
     method,
-    headers: { accept: "application/json" },
+    headers: { accept: "application/json", ...auth },
     cache: "no-store",
   };
   if (body !== undefined) {