- read [ignored_tags] instead of [ignored_domains], so tag-based ignoring
actually works
- match ignored tags as whole tags, not substrings ('ipt' no longer matches
'iptorrents')
- send the login-error notification through the Pushover client instead of
the non-existent module-level send_message()
- handle an empty torrent list instead of raising IndexError before any
summary is sent
- tag a new torrent once instead of once per code path
- honour toml_path without DOCKER in AppriseClient
- close replaced log handlers so the file descriptor is released
Also add the missing [telemetry] section to config.toml.example, which
qbit-maid.py requires at startup.
The tests that pinned each bug now assert the fixed behaviour, and the
suite still reports 100% coverage.
The runner image ships pip 24.0 from the Debian package, so
`pip install --upgrade pip` aborts with "Cannot uninstall pip 24.0,
RECORD file not found" and the job failed after 4s. Drop the upgrade
and install with --break-system-packages.
The suite had 8 empty `pass` tests, two tests that made live network
calls to a production healthcheck endpoint, and a write_csv test that
only asserted the file existed. CI ran only test_komodo_deploy, so the
main suite never executed at all.
- Split tests by module: test_qlist, test_qprocess, test_qlogging_extra,
test_apprise, test_qbt; rewrote test_qbitmaid and test_write_csv.
- qbt_fakes.py: shared fakes (FakeQbt, FakeQbtClient, FakeAPIError,
StubPushover, FailingPushover, torrent factory).
- test_qbt.py loads the hyphenated qbit-maid.py via importlib and covers
the __main__ entry point.
- Replaced the live-network tests with a RecordingRequests stub
(suite time 14s -> 0.035s).
- .coveragerc + CI unit-tests job running the full suite with
--fail-under=100.
127 unit tests, 100% statement coverage on all 5 first-party modules
(440 stmts), 15/15 mutations caught.
Seven pre-existing bugs are pinned by tests with KNOWN BUG docstrings
and left unfixed pending a decision.
Switch the summary to the layout jblu picked: each counter renders as
"emoji value label", grouped two per line and joined by a middle dot.
Scanned 42 torrents in 12s
🗑️ 7 removed · 👻 4 orphaned
⏭️ 5 ignored · 🐣 3 premature
🛡️ 10 protected · 🌐 20 public
- the value leads, so the numbers are the first thing the eye lands on
- html style now bolds the values instead of the labels
- "Non-protected" reads as "public", matching the non_protected_tag
- use_emoji now defaults to true; false gives the plain variant
- dry runs read "7 would be removed"
Pushover renders the notification body in a proportional font, so the
space-padded label/value columns never lined up on a phone — the body
collapsed into a ragged flat list with no visual grouping. The html
variant was worse: the padding sat inside the HTML, so it collapsed to a
single space.
Counters are now grouped two per line and joined with a middle dot, which
survives any font. The body drops from 8 lines to 5 and no longer restates
the outcome the title already carries.
- replace _notify_row/LABEL_WIDTH/VALUE_WIDTH with _stat_row
- add docs/notify-preview.py, which renders the real build_summary()
output as a phone-width PNG so layout changes can be reviewed visually
- tests: assert no space padding, two counters per row, and that the body
does not repeat the title
Rework the Pushover/Apprise summary so it reads as a report instead of a
dump of counters.
- Extract build_summary() as the single source of truth for both notifiers,
removing the duplicated body string and the stray leading indentation
caused by backslash line continuations.
- Add a 'format' option: 'plain' (aligned columns) or 'html' (bold labels,
sets html=1 for Pushover). Apprise always receives the plain body.
- Add 'use_emoji' for emoji-prefixed rows.
- Title now states the outcome ('7 torrents removed') instead of
'--- qbit-maid summary ---'.
- Say 'Would remove' when delete_torrents is false, so a dry run is never
mistaken for a real deletion.
- Format runtime as 12s / 1m 04s instead of a timedelta with microseconds.
- Add include_hostname, notify_on_empty, priority, sound, retry and expire.
- Log notification failures instead of letting them abort the run.
- Cover the new formatting with unit tests; document the options in the
README and config.toml.example.
':?' exits 1 under bash but 2 under dash (the Ubuntu runner's sh),
so the two missing-variable tests failed in CI while passing locally.
Replace with explicit checks so every failure path exits 1.
15 unittest cases exercising the real script end to end: target
selection, auth headers, request shape, success/failure, HTTP 401/400,
unreachable host, polling to completion, and timeout.
Runs in CI via .gitea/workflows/test.yml on push to main/dev* and on
pull requests. No network or live Komodo instance required.
core.fileMode is false in this repo, so the exec bit on
komodo-deploy.sh was not recorded and the runner hit
'Permission denied'. Mark the file executable in the index and
invoke it via 'sh' so the step no longer depends on the mode bit.
Adds a shared .gitea/scripts/komodo-deploy.sh that calls the Komodo
/execute/DeployStack API after the image is pushed, then polls
/read/GetUpdate until the deploy completes.
A push webhook cannot do this correctly: it fires before the image
exists in the registry, so Komodo would redeploy the previous :latest.
- docker.yml deploys stack 'qbit-maid' (required)
- docker-dev.yml deploys vars.KOMODO_STACK_DEV, skipped when unset
- Add setup_logging() returning a named 'qbit-maid' logger with
RotatingFileHandler (configurable size/backups) and a console handler
- Add log_torrent() helper for structured torrent context
- Replace basicConfig-based tor_log() with setup_logging()
- Remove redundant 'if self.use_log:' guards across qbit-maid.py,
qlist.py, qprocess.py, qlogging.py (level filtering handles it)
- Document new logging options in config.toml.example and README
trex Bot
jblu
nanobot
jblu