Honest list of what version 0.1.0 does not do, and behavioural constraints that surprise people. Feature requests belong in the project tracker, not in .env.
- No
/send, ad-hoc page composer, or club broadcast.
- No POCSAG listener (the bot does not receive pages).
- No configurable page templates (format is fixed, 80 characters).
- No extra “Telegram only” reminder channel; DAPNET and Telegram share the same offsets.
- SQLite only — no PostgreSQL, MySQL, or Redis.
- No CalDAV push / webhooks — calendars are polled (
SYNC_INTERVAL_SECONDS, default 10 minutes).
- No Alembic — schema changes are
create_all plus a couple of ALTER TABLEs.
- No i18n — Telegram UI is English only.
- No Telegram webhooks — long polling only.
- No group-chat workflow, topics, or channel posting.
- You cannot address a single RIC when several pagers share one callsign.
POST /calls uses callSignNames; DAPNET delivers to all pagers of that callsign.
- There is no “send to this RIC number” field. The bot’s RIC list is a map onto callsigns.
- An invalid transmitter group name can cause DAPNET to drop the whole call.
- User lookup during
/register uses DAPNET_DEFAULT_SERVER (usually Germany), not the Spain/Germany button the user just tapped for sending. A callsign that exists only on the other core may fail registration.
- The
pagers array on GET /callSigns/{cs} is often hidden for non-admin operators. Existence of the user + subscriber is then enough; RadioID and manual RIC entry identify pagers.
- API 2.0 is not supported.
- If DAPNET fails but Telegram succeeds, that offset is marked sent — no DAPNET retry.
- After a process crash, only reminders still inside
MISFIRE_GRACE_MINUTES (default 30) are sent. Longer outages skip old due times on purpose (no flood).
- JobQueue is in-memory; restart does not restore “next run in 3 minutes” — intervals start again after 10/20 seconds.
- Events created after a lead offset has passed will not get that offset (the fire time is already in the past and may be outside grace). Later offsets can still fire.
- Recurrence expansion depends on python-caldav 3.x
search(..., expand=True). Odd CalDAV servers may return unexpanded or incomplete series.
- All-day events always use 09:00 local; you cannot choose another hour.
- Lead times are a fixed set (15 / 30 / 60 / 180 / 1440 / 10080 minutes), not an arbitrary number.
/refresh lists at most 15 upcoming events in the reply; more still sit in the cache.
/users lists at most 50 people.- RIC picker buttons show at most 12 rows; TX group buttons at most 18 names (commands still work).
- One Telegram user id → one bot account (no multiple Nextcloud profiles per Telegram user).
- Registration and add-RIC conversations time out after 600 seconds.
- Background CalDAV errors notify the user in Telegram; they do not page the operator.
- No dashboard, metrics, or health HTTP endpoint.
- No automatic database backup.
- Callsign pattern is
3–20 of [a-z0-9/] — unusual callsign punctuation is rejected.
- Fernet key loss = stored Nextcloud passwords unreadable.
- Admins see Nextcloud URL + username on signup (not the password).
- Plain-text Telegram (by design).
- SQLite file is the entire state; concurrent hosts pointing at the same DB file over NFS are unsupported.