Review fixes for the persistent logs:
- each service writes its own folder (the old-file cleanup matches by
prefix, so 'worker' deleted the 'worker-ai' files);
- at most LOG_MAX_DAY_MB (50) a day, and at least 2 days kept;
- written by a background thread, not on the async workers;
- e-mail addresses and tokens in request paths are masked;
- a failing health check is logged again, a passing one is not;
- one-off commands (compare-models) log to the console only.
On a stop the worker takes nothing new and lets running work finish
(SHUTDOWN_DRAIN_SECS, worker 40 s, worker-ai 270 s, with matching stop
grace periods), so a deploy no longer cuts an analysis off, wastes its
model call and delays it by the 15-minute stuck-job requeue.
The CI cleanup caps the build cache at 10 GB: by age it had grown to
33 GB on the runner and filled its disk.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL
Backend and the workers also write their log as JSON into the logs
volume, one file a day kept 30 days (LOG_DIR, LOG_NAME, LOG_KEEP_DAYS),
so errors can be looked up after a container is replaced. The Docker
health check (/healthz every 30 s) no longer goes into the request log
of the backend and the horoscope API.
The home, login and packages pages no longer compare the analyses to
an experienced astrologer's depth; they say what the user gets:
detailed, personal text in minutes, for a fraction of a personal
consultation's price.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL
A paid checkout queues an invoice row with every invoice detail (buyer,
tax type and number, item, net, VAT, gross). The worker's new 'invoices'
task creates the Billingo partner and the paid card invoice and e-mails
it (API v3), saving each step so a retry never issues a second invoice,
and keeps Billingo's whole answer; the PDF is fetched from Billingo when
the user downloads it from the new Számláim page (in the menu once there
is an invoice). Payments and invoices outlive a deleted account.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL
The catalog stays in packages.rs: every checkout session carries the
package's name and gross price, nothing has to be kept in Stripe. The
package is granted once, when Stripe confirms the payment: by the signed
webhook, or by the return page asking Stripe when the webhook is late.
With a test key only admins can buy, so Stripe's test cards give nobody
free packages. Without the keys nothing changes.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL
WORKER_TASKS (email, cleanup, analyses; all by default) picks the jobs of a
worker process. In production the worker sends e-mails and cleans up (and
will take the other background jobs), worker-ai writes the analyses and
answers with its quick and slow lanes, sized from the server's .env.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL
main builds images tagged with the commit hash and `main`; prod also tags
`prod` and `latest` and deploys. A prod run without DEPLOY_HOST now fails
instead of skipping silently.
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL
- .gitea/workflows/ci.yml: test (Postgres service), build and push the
four images tagged with the commit hash, deploy over SSH when
DEPLOY_HOST is set
- ci/runner: act_runner 0.6.1 on the dev server using the host's Docker
- deploy/: production compose with Caddy (https), daily pg_dump backups,
log rotation; setup guide
Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_012vLHzh5qUhFtmJhzkcTbXL