Logto's OIDC endpoint may respond before the Management API is fully
initialized. Add a retry loop that checks GET /api/roles returns valid
JSON before making any API calls. Fixes intermittent bootstrap failure
on cold starts with 'Cannot index string with string "name"'.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
1. Docker socket security: remove root group from Dockerfile, use
group_add in docker-compose.yml for runtime-only socket access
2. M2M server communication: create ServerApiClient using Logto
client_credentials grant with API resource scope. Add M2M server
role in bootstrap. Replace hacky admin/admin login in
AgentStatusService.
3. Async deployment: extract DeploymentExecutor as separate @Service
so Spring's @Async proxy works (self-invocation bypasses proxy).
Deploy now returns immediately, health check runs in background.
4. Bootstrap: M2M server role (cameleer-m2m-server) with server:admin
scope, idempotent creation outside the M2M app creation block.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Check for spaClientId and m2mClientSecret in the cached bootstrap
file. If both exist, exit immediately instead of re-running all
phases. Delete /data/logto-bootstrap.json to force a re-run.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Change rolesClaim from "scope" to "roles" to match the custom claim
injected by the Logto Custom JWT script
- Add Phase 7b: configure Logto Custom JWT for access tokens that maps
org roles (admin→server:admin, member→server:viewer) and global roles
(platform-admin→server:admin) into a standard "roles" claim
- Add additionalScopes field to OIDC config
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ADMIN_ENDPOINT is now HTTPS so admin-tenant calls need the
forwarded proto header. Default-tenant calls stay unchanged.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Revert all Traefik port 3002 and ADMIN_ENDPOINT changes that broke
bootstrap. Admin console HTTPS access needs a different approach.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ADMIN_ENDPOINT is http://localhost:3002, but bootstrap sent
Host: PUBLIC_HOST:3002 which didn't match. Let curl use the
default Host from LOGTO_ADMIN_ENDPOINT (logto:3002) which Logto
resolves to the admin tenant internally.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Default tenant (port 3001) works without it — adding it caused
Internal server error. Only the admin tenant needs it because
ADMIN_ENDPOINT changed to HTTPS.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All Logto endpoints are configured with HTTPS but bootstrap calls
internal HTTP. Every curl call needs the forwarded proto header.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Logto's ADMIN_ENDPOINT is now HTTPS but bootstrap calls the internal
HTTP endpoint directly. TRUST_PROXY_HEADER needs X-Forwarded-Proto
to resolve the correct scheme.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Admin tenant defaults to Register mode (onboarding flow). Since we
create the admin user via API, we need to switch to SignIn mode so
the custom sign-in UI can authenticate against the admin console.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The admin tenant requires both the 'user' role (base access) and
'default:admin' role (Management API). Missing the 'user' role
causes a 403 at the identification step.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The admin console runs on a separate tenant with its own user store.
Previous approach tried to assign a non-existent 'admin:admin' role
on the default tenant. Now creates the user on the admin tenant via
port 3002, assigns 'default:admin' role for Management API access,
and adds to t-default organization.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Change chmod 600 to 644 on bootstrap JSON (cameleer user needs read)
- Use PNG favicon instead of SVG (currentColor invisible in browser tab)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
The server-ui logout redirects to /server/login?local but this URI was
not whitelisted in Logto, causing the post-logout redirect to fail.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- Add server:admin/operator/viewer scopes to bootstrap and org roles
- Grant SaaS admin Logto console access via admin:admin role
- Configure sign-in experience with Cameleer branding (colors + logos)
- Add rolesClaim and audience to server OIDC config
- Add server scopes to PublicConfigController for token inclusion
- Permit logo SVGs in SecurityConfig (fix 401 on /platform/logo.svg)
- Add cameleer3 logo SVGs (light + dark) to ui/public/
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Without this, Logto returns consent_required when the server tries
SSO because the scopes were never explicitly granted.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Server OIDC callback is at /oidc/callback (without /server/ prefix due
to strip-prefix). Register both variants until server reads
X-Forwarded-Prefix.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Move SPA assets from /assets/ to /_app/ (Vite assetsDir config) so
Traefik can route /assets/* to Logto without conflict. All services
on one hostname with path-based routing:
- /oidc/*, /interaction/*, /assets/* → Logto
- /server/* → server-ui (prefix stripped)
- /api/* → cameleer-saas
- /* (catch-all) → cameleer-saas SPA
Customer needs only 1 DNS record. Server gets OIDC_JWK_SET_URI for
Docker-internal JWK fetch (standard Spring split config).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Standard OIDC architecture: subdomain routing (auth.HOST, server.HOST),
TLS via Traefik, self-signed cert auto-generated on first boot.
- Add traefik-certs init container (generates wildcard self-signed cert)
- Enable TLS on all Traefik routers (websecure entrypoint)
- HTTP→HTTPS redirect in traefik.yml
- Host-based routing for all services (no more path conflicts)
- PUBLIC_PROTOCOL env var (https default, configurable)
- Protocol-aware redirect URIs in bootstrap
- Protocol-aware UI fallbacks
Customer bootstrap: set PUBLIC_HOST + DNS records + docker compose up.
For production TLS, configure Traefik ACME (Let's Encrypt).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Logto's login page references /assets/* which conflicts with the SPA's
assets at the same path. Using Host-based routing (auth.localhost) gives
Logto its own namespace - all paths on that subdomain go to Logto,
eliminating the conflict. *.localhost resolves to 127.0.0.1 and is a
secure context.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Logto ENDPOINT now points at Traefik (http://localhost) instead of
directly at port 3001. All services share the same base URL, eliminating
OIDC issuer mismatches and crypto.subtle secure context issues.
- Remove :3001 from all public-facing Logto URLs
- Add cameleer3-server-ui to Traefik at /server/ with prefix strip
- Dashboard link uses /server/ path instead of port 8082
- Bootstrap Host headers match Logto ENDPOINT (no port)
- Redirect URIs simplified (Traefik handles port 80)
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
All public-facing URLs (Logto OIDC, redirect URIs, dashboard links) now
derive from PUBLIC_HOST in .env instead of scattered localhost references.
Resolves Docker networking ambiguity where localhost inside containers
doesn't reach the host machine.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Bootstrap was sending LOGTO_PUBLIC_ENDPOINT (http://localhost:3001)
as the OIDC issuer URI to the server. Inside Docker, localhost is
unreachable. Changed to LOGTO_ENDPOINT (http://logto:3001).
Also: .env must set LOGTO_ISSUER_URI=http://logto:3001/oidc (not
localhost) since this env var feeds cameleer3-server's OIDC decoder.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Phase 7 server health check failed intermittently due to timing.
Add 3-attempt retry loop with 2s sleep. Guard against jq returning
literal "null" string for TRAD_SECRET. Add debug logging for both
preconditions.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
ConnectivityHealthCheck: /actuator/health → /api/v1/health (actuator
now requires auth on cameleer3-server after OIDC was added).
Bootstrap: POST → PUT for /api/v1/admin/oidc (server expects PUT,
POST returned 405 causing OIDC config to silently fail).
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Add @PreAuthorize annotations to all API controllers (14 endpoints
across 6 controllers) enforcing OAuth2 scopes: apps:manage, apps:deploy,
billing:manage, observe:read, platform:admin.
Enforce tenant isolation: TenantResolutionFilter now rejects cross-tenant
access on /api/tenants/{id}/* paths. New TenantOwnershipValidator checks
environment/app ownership for paths without tenantId. Platform admins
bypass both layers.
Fix frontend: OrgResolver split into two useEffect hooks so scopes
refresh on org switch. Scopes now served from /api/config (single source
of truth). Bootstrap cleaned — standalone org permissions removed.
Update docs/architecture.md, docs/user-manual.md, and CLAUDE.md to
reflect all auth hardening changes.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Creates 10 API resource scopes (platform:admin + 9 tenant-level) on the
Cameleer SaaS API resource, assigns all to platform-admin role, creates
matching organization scopes, and wires them declaratively to org roles
(admin gets all, member gets deploy/observe). All operations are idempotent.
Co-Authored-By: Claude Sonnet 4.6 <noreply@anthropic.com>
The bootstrap script runs before cameleer-saas (Flyway), so tenant
tables don't exist yet. Moved DB seeding to BootstrapDataSeeder
ApplicationRunner which runs after Flyway migrations complete.
Reads bootstrap JSON and creates tenant/environment/license if missing.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Bootstrap script now creates:
- SaaS Owner (admin/admin) with platform-admin role
- Tenant Admin (camel/camel) in Example Tenant org
- Traditional Web App for cameleer3-server OIDC
- DB records: tenant, default environment, license
- Configures cameleer3-server OIDC via its admin API
All credentials configurable via env vars.
Backend:
- Fix LogtoManagementClient resource URL (https://default.logto.app/api)
- Add getUserRoles/getUserOrganizations to LogtoManagementClient
- Add GET /api/me endpoint (user info, platform admin status, tenants)
- Add GET /api/tenants list-all for platform admins
- Remove insecure X-header forwarding from Traefik
Frontend:
- Org-scoped tokens: getAccessToken(resource, orgId) for tenant context
- OrgResolver component populates org store from /api/me
- useOrganization Zustand store (currentOrgId + currentTenantId)
- Platform admin sidebar section + AdminTenantsPage
- View Dashboard link points to cameleer3-server on port 8081
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Logto returns opaque access tokens when no resource is specified.
Added API resource creation to bootstrap, included resource indicator
in /api/config, and SPA now passes resource parameter in auth request.
Also fixed issuer-uri to match Logto's public endpoint.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Separate LOGTO_PUBLIC_ENDPOINT (browser-facing, defaults to
http://localhost:3001) from LOGTO_ENDPOINT (Docker-internal).
Also fix bootstrap M2M verification by using correct Host header
for default tenant token endpoint.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
Logto routes requests by Host header to determine tenant. Inside Docker,
requests to logto:3001/3002 need Host: localhost:3001/3002 to match the
configured ENDPOINT/ADMIN_ENDPOINT.
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
- logto-bootstrap.sh: API-driven init script that creates SPA app,
M2M app, and default user (camel/camel) via Logto Management API.
Reads m-default secret from DB, then removes seeded apps with
known secrets (security hardening). Idempotent.
- PublicConfigController: /api/config public endpoint serves Logto
client ID from bootstrap output file (runtime, not build-time)
- Frontend: LoginPage + CallbackPage fetch config from /api/config
instead of import.meta.env (fixes Vite build-time baking issue)
- Docker Compose: logto-bootstrap init service with health-gated
dependency chain, shared volume for bootstrap config
- SecurityConfig: permit /api/config without auth
Flow: docker compose up → bootstrap creates apps/user → SPA fetches
config → login page shows → sign in with Logto → camel/camel
Co-Authored-By: Claude Opus 4.6 (1M context) <noreply@anthropic.com>
