Fizzy is now free. Remove the entire Stripe billing system,
subscription management, and card/storage limit enforcement.
Removes from saas/: Plan model, Account::Billing, Account::Subscription,
Account::Limited, Account::OverriddenLimits, Account::BillingWaiver,
all subscription/billing controllers and views, Stripe webhook handler,
card creation/publishing limit enforcement, admin account override UI,
usage report rake task, and all related tests.
Removes from main app: Fizzy.saas? guards for subscription panel,
SaaS card footer override, near-limit notices, and saas.css stylesheet.
Adds migration to drop billing tables from the SaaS database.
Non-billing SaaS features (push notifications, signup, authorization,
telemetry, console1984/audits1984) are preserved.
* Add saas:usage_report rake task and extract Subscription.paid scope
Add a rake task to generate a CSV usage report with per-account data:
Queenbee ID, sign up date, paid date, card count, storage used, and
last active date.
Extract the paid subscriptions query from Admin::StatsController into
an Account::Subscription.paid scope so both the controller and the
new rake task can share it.
* Add comped and account name columns to usage report
* Update saas/lib/tasks/fizzy/usage_report.rake
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Preload storage_total to avoid N+1 in usage report
* Batch last_active_at queries to avoid per-account aggregates
* Add tests for Account::Subscription.paid scope
* Update saas/app/models/account/subscription.rb
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
* Aggregate paid dates in SQL instead of loading all subscriptions
* Fix paid scope to derive plan keys from Plan.all instead of PLANS hash
* Move paid dates and comped lookups into per-batch queries
* Materialize batch IDs to avoid cross-database subquery
SaasRecord models live on a separate database (fizzy_saas) that doesn't
have the accounts table. Using batch.select(:id) generated a subquery
that ran on the saas database, causing a table-not-found error. Using
pluck(:id) materializes the IDs into an array instead.
---------
Co-authored-by: Copilot <175728472+Copilot@users.noreply.github.com>
The True-Client-IP header is set by Cloudflare and is only trustworthy
when behind a Cloudflare proxy. In non-Cloudflare deployments, this
header is attacker-controlled and can be used to spoof IP addresses.
Moving the middleware into the saas engine ensures it only loads for our
Cloudflare-fronted production deployment, not for self-hosted OSS
instances.
GHSA-cpch-9qg2-x8fq
Use ActionPushNative's new on_load hook to configure the database connection,
following the same pattern as Active Storage and Action Text:
ActiveSupport.on_load(:action_push_native_record) do
connects_to database: { writing: :saas, reading: :saas }
end
This allows ApplicationPushDevice to inherit directly from ActionPushNative::Device
without needing an intermediate abstract class.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Add disallow_account_scope to skip tenant requirement
- Move routes to saas/config/routes.rb (engine routes)
- Use saas.devices_path/saas.device_path for engine route helpers
- Update tests to work without tenant context
Devices belong to Identity (global), not Account, so they don't
need tenant context in the URL.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
When a session is destroyed (user logs out), all devices registered
to that session are automatically destroyed, preventing push
notifications from being sent to logged-out devices.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Replace NotificationPusher with a cleaner architecture:
- Add Notification::Pushable concern with push target registry
- Add Notification::Push base class with template methods
- Add Notification::Push::Web for web push (OSS)
- Add Notification::Push::Native for native push (SaaS)
- Add Notification::WebPushJob and Notification::NativePushJob
Key design:
- Registry pattern: Notification.register_push_target(:web)
- Template method: push calls should_push? then perform_push
- Subclasses override should_push? (with super) and perform_push
- Each target handles its own job enqueueing
Also:
- Add Notification#pushable? for checking push eligibility
- Add Notification#identity delegation to user
- Reorganize tests to match new class structure
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Tidy up saas engine a bit more
Devices now belong to Identity instead of User, allowing a single
device registration to work across all accounts an identity has
access to.
- Move User::Devices to Identity::Devices
- Update DevicesController to use Current.identity
- Update NotificationPusher::Native to use user.identity.devices
- Clean up tests to use @identity directly
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
Fix reference to `user.devices`, left-over from the identity switch
- Use RESTful DELETE /devices/:id where :id can be token or database ID
- Remove redundant unregister collection route
- Remove old Users::DevicesController
- Return 404 when device not found instead of silently succeeding
- Return 422 for invalid platform via ActiveRecord validation
- Update action_push_native to main branch (includes validate: true on enum)
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove Users namespace from DevicesController (now just DevicesController)
- Create ApplicationPushDevice model extending ActionPushNative::Device
- Move device registration logic (find_or_initialize + update) to model
- Update User::Devices concern to use ApplicationPushDevice
- Fix push notification tests (endpoint validation, job count expectations)
- Update push_config_test to use ActionPushNative.config
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Remove UUID column from devices table entirely
- Update unique index from (owner, uuid) to (owner, token)
- Simplify create action to just create device records
- Add token-based unregister route for API clients
- Consolidate error handling with rescue_from
- Update fixtures to remove uuid references
Devices are now identified by (owner, token) instead of UUID.
This simplifies the client-side registration flow.
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
- Fix owner_id type to UUID in devices migration
- Fix NOT NULL crash in device registration
- Add APNS config and 1Password integration
- Add --apns flag to bin/dev for local development
- Clean up devices controller
Rename histogram from :request_wait_seconds to :request_wait so Yabeda's
unit: :seconds produces gvl_request_wait_seconds instead of
gvl_request_wait_seconds_seconds.
* Add GVL contention metrics via gvltools and Yabeda
Expose per-request and process-wide GVL wait time as Prometheus metrics
to diagnose suspected GVL contention from Solid Cable and Action Cable
threads in Puma workers.
Metrics: gvl_request_wait_seconds (histogram), gvl_waiting_threads
(gauge), gvl_global_timer_total_seconds (gauge).
* Add unit: :seconds to GVL histogram for consistency
- Rename SaasAdminController to Admin::AuditController
- Override require_authentication to support bearer tokens alongside
session auth, allowing API access to audit console
- Add migration for audits1984_auditor_tokens table
- Add controller tests for authentication scenarios including staff
validation on token-based access
Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
* Sync owner email to Stripe when changed
When an account owner changes their email address, update the
corresponding Stripe customer record via a background job.
Also handles ownership transfers: when a user becomes the account
owner, their email is synced to Stripe.
Responsibility chain:
- User::NotifiesAccountOfEmailChange triggers on owner identity change
or when a user becomes owner
- Account::Billing#owner_email_changed enqueues sync job
- Account::SyncStripeCustomerEmailJob performs the update with
polynomial backoff retries
- Account::Subscription#sync_customer_email_to_stripe calls Stripe API
* Address PR feedback: error handling and test coverage
- Handle Stripe::InvalidRequestError in sync_customer_email_to_stripe
(mirrors cancel method behavior for deleted customers)
- Add test for deactivated owner (owner with nil identity)
- Add test for deleted Stripe customer scenario
Before, we were relying on just changing the cards_count in account, but this
could create problems where the system to calculate the next card number fails due
to the unique constraint.
Instead:
- Always let you create drafts when pressing "Add card"
- Prevent creations of published cards via API
- Prevent publication of cards in all cases when the limit is exceeded
* Sentry: use KAMAL_VERSION for release tracking
Kamal injects KAMAL_VERSION at container runtime with the git SHA,
so no need to bake it into the Docker image via build args.
* Sentry: serve as Rails error reporter
Sentry receives errors from Rails.error.report and Active Job
`retry_on/discard_on report: true`.
Error context is provided by Rails.error.set_context in Fizzy's
Authentication concern and ApplicationJob.
To keep as much of this as we can in the `fizzy-saas` gem, this PR
opts to store console/audits models in a separate database, and so:
- the gem contains the migrations and the database config
- the app contains a separate schema file distinctly for SaaS concerns
- Rails' current behavior prevents us from easily keeping this file in the gem
Also note that the stock `audits1984` schema is updated to reference
the auditor via a UUID foreign key.
Finally, in order for the database schema file to be located in the
gem directory (and not the application directory), it's necessary to
monkeypatch `AR::DatabaseTasks.schema_dump_path` to support absolute
paths. This functionality has been proposed upstream in
https://github.com/rails/rails/pull/56290 but is awaiting a decision
from the core team.
ref: https://app.fizzy.do/5986089/cards/2469
ref: #17
Jorge Manrubia