v1.4.0

Highlights

• The new Activity page shows active and completed plays, source delays, provider contributions, recovery activity, and Plex playback context without treating successful fallback as a viewing failure.
• Playback now detects silent or truncated article bodies, retries stalled sequential and pipelined reads on healthy connections, and protects provider capacity for playback while preparation or health checks are busy.

Fixes

• Troublesome NZBs now receive a bounded provider-coverage probe before full verification. Releases that cannot be verified fail promptly with the full reason and time-to-fail visible in Watchdog, while healthy providers retain the fast bulk path.
• Preparation skips files that cannot contribute useful media or recovery metadata, reducing unnecessary first-article work and avoiding delays from irrelevant payloads.
• Health checks preserve completed work when a pipelined provider stalls, grow useful provider capacity on demand, and avoid assigning normal bulk work to providers that failed qualification.
• Multi-volume and season-pack RAR preparation now handles boundary volumes and sparse metadata more reliably.
• Runtime dependencies now use patched releases for recently disclosed web-server, archive, API-schema, and SQLite security issues.

Improvements

• Read-ahead is now configured in MiB instead of article count, providing predictable buffering and memory use across NZBs with different article sizes while automatically adapting the number of prefetched articles per stream.
• Watchdog includes provider probing in the Health and total timings, keeps detailed probe and bulk measurements in the expanded statistics, and presents failure reasons without requiring the statistics panel.
• Queue, Watchdog, activity navigation, and playback status use denser layouts that remain readable on narrow screens.
• Plex connection testing and playback attribution provide clearer evidence about scans, imports, direct play, transcoding, and viewer impact.

v1.3.20

1.3.20 (2026-07-25)

Improvements

• The Watchdog page now shows each playback attempt in far less space. The counters above the filters are gone, since the filter buttons already carry the same totals, and a run that only had one attempt no longer draws a table that repeats what the card header already says.
• Resolved and failed runs are marked with a check or a cross instead of a word, and the release size moved up beside the category so the summary line carries only timings.
• Release names that no longer fit are wrapped onto further lines instead of being cut off, on both narrow screens and full-width layouts.
• The button that opens run statistics is now a distinct control at the end of the summary line, and it stays on that line instead of dropping below the timings.
• Repeated details were removed throughout: the indexer is only named when one is known, a candidate is only listed when it differs from the release named in the card header, and the indexer and reason columns appear only when a run actually recorded them.
• On phones, run statistics are easier to read. Each figure is labelled above its value rather than beside it, and long summaries wrap by phrase instead of running together. Per-provider working time and per-provider rate are left out of the narrow layout, since both can be read from the values still shown.
• Provider shares are no longer printed on every row of the main view. Provider names remain, with the shares available in the provider popover and in the routing tables under run statistics.

v1.3.19

1.3.19 (2026-07-24)

Fixes

• Background health checks can no longer be held up by a single file. A file that could not be scheduled, whose first article was already gone, or whose check ended without a verdict is now given a verdict or a retry time instead of being retried continuously ahead of everything else.
• File checks no longer pause near completion when a provider stops responding part-way through a batch. The stalled connection is dropped after two seconds and only its unanswered articles move to another provider, instead of waiting for that provider to close the connection.
• A provider that loses a single connection is no longer set aside for the rest of a check. Only a provider that never answers is, so its remaining connections keep doing useful work.
• Large releases now get verification time proportional to their size. Previously a fixed budget could expire on every attempt, leaving those files reported as unverified after every retry.

Improvements

• Health checks no longer pause connection upkeep on providers they are not using, so playback and preparation keep their ready connections while a check runs.

Fixes

v1.3.18

Fixes

• Health checks now preserve pipelined article-status replies as they arrive and retire a connection if a partial batch is abandoned. Slow or interrupted batches can retain their successful work without leaving unread replies to interfere with later checks.

Improvements

• Playback logs now correlate startup, seeks, first-byte timing, upstream and downstream stalls, connection waits, provider rotations, and backup-provider recovery in one request summary. This makes it possible to distinguish Usenet trouble from a slow player, proxy, or mobile connection.
• History can now be filtered by category. On-demand movie and series playback is shown separately from normal automation imports while retaining the physical categories expected by streaming integrations.

Required AIOStreams setup

• In the AIOStreams NzbDav service settings, enter the NzbDavEx API key as streaming:<NZBDAV API key>. Keep the API key stored in NzbDavEx unchanged; the prefix is added only in AIOStreams so NzbDavEx can identify and categorize its streaming submissions.

v1.3.17

Health checks now detect connected provider sockets that stop returning STAT responses and promptly retry only unanswered articles on another provider. This prevents an isolated stalled batch from holding the queue at 99% while preserving full article verification and treating provider outages as unverifiable rather than missing.

v1.3.16

This release replaces the fragile preparation and health-check fallback behavior in 1.3.15 with bounded, evidence-based provider routing. Normal complete releases retain the fast parallel path, while incomplete releases can consult backup providers without leaving queue work pinned indefinitely or treating an unavailable provider as proof that an article is missing.

Fixes

• Bulk health checks now distribute primary lanes using established usable connections, keep failed probes out of normal batch chains, and let cold providers join after a successful late probe. Deferred checks refresh their warm capacity before starting, and foreground demand can continue growing beyond the initial ready-connection floor.
• Article verdicts now distinguish confirmed missing from temporarily unverifiable. Missing is reported only after every eligible provider answers; unavailable providers produce a bounded retryable result and cannot trigger an incorrect repair or poison the missing-article cache.
• Unresolved health batches receive one coordinated, time-bounded recovery pass. Cold plain Backup providers enter that pass directly instead of first burning a normal lane timeout, and their answers are included accurately in Watchdog provider statistics.
• First-segment preparation now uses one bounded provider pass with per-host backup admission, availability probing, and separate connection-acquisition and NNTP-command deadlines. Missing articles fail clearly, provider outages remain distinguishable, and a bad NZB no longer cycles indefinitely in the queue.
• The confirmed-missing cache now expires entries after six hours, caps itself at 100,000 segments, and still clears immediately when provider configuration changes.

Improvements

• Queue rows show a restrained live notice only while preparation or health checking is actively consulting fallback providers. Completed recovery and failure details remain in the persistent Watchdog report.
• Watchdog records the full overlapped health runtime separately from the time health checking blocks after preparation. Compact cards show Prep, blocking Health, and their press-to-play Total; expanded statistics retain both health measurements and detailed preparation attempts, misses, unanswered checks, work time, and data use per provider.
• Existing providers can be saved without opening a test connection when their host, port, SSL mode, username, and password are unchanged. Test Connection remains available as a separate action and is still required for new or changed connection details.

Notes

• A backup-heavy or genuinely incomplete NZB can take longer than a complete release because nzbdavex deliberately asks additional eligible providers before declaring an article missing. That search is bounded and its active fallback state is visible in the queue.
• A queue item can remain visible at 100% briefly after health checking finishes while nzbdavex creates WebDAV mount metadata and saves the import to its database. This local import phase is not a stalled provider health check.

v1.3.15

Fixes

• Connection pools now continue bounded authentication handshakes while foreground work reuses returned idle sockets. Health checks can therefore grow rapidly beyond their initial warm connection count without trapping active work behind connection setup.
• Health-check STAT routing now accounts for queued acquisitions and separates connection-acquisition deadlines from command deadlines. Congested providers fail over promptly, newly authenticated sockets receive a full command window, and unresolved batches receive one globally bounded recovery pass.
• Queue retries now wake at their advertised retry time instead of waiting for the next one-minute poll.
• Transient first-segment failures retry only the affected file instead of restarting all preparation work.

Notes

A queue item can remain visible at 100% briefly after health checking finishes while nzbdavex creates WebDAV mount metadata and saves the import to its database. This local import phase is not a stalled provider health check.

Validation included 309 passing backend tests, a clean v1.3.15 release-container build, and live health-check runs against the configured providers.

v1.3.14

Improvements

• The Usenet provider overview now groups providers by role, shows pipeline overrides directly on each card, and keeps drag-and-drop ordering within the relevant provider group.

Fixes

• Queue and health-check prewarming now establishes missing connections without borrowing authenticated idle sockets for a simultaneous validation burst, keeping foreground preparation and health work supplied.
• Ready-connection maintenance now refreshes the full idle target in rolling batches of four and runs only while the provider has no active or pending foreground work.
• Health-check STAT priming now samples at most four sockets per provider. Provider timeouts remain retryable, and canceled warmup cleanup cannot hold the queue indefinitely.

Notes

• A queue item can remain visible at 100% briefly after its health check has completed while nzbdavex creates the WebDAV mount metadata and saves the final import to its database. This is local import work, not a stalled provider health check; import progress is not currently displayed separately.

v1.3.13

Fixes

• Health checks now keep primary article checks on providers with established connections instead of reserving work for providers with no usable sockets, preventing checks from lingering at 99%.
• Stalled health-check batches now retry only unresolved articles with another provider and fail within a bounded time if no provider completes them.

v1.3.12

Improvements

• Provider passwords can now be shown or hidden from inside the password field, making saved credentials easier to verify while keeping them hidden whenever the provider editor is opened.