Write per-track match diagnostics to a file instead of relying on HA's log #55

Merged

neerdoc merged 1 commit from macsome_14 into main

2026-08-02 21:22:36 +00:00

neerdoc commented

2026-08-02 21:22:33 +00:00

Owner

_plex_find() now writes a full diagnostic line per track to
match_debug.log, created next to engine.py in the installed integration
folder (not subject to HA's log rotation or level config), truncated at
the start of every match_and_build() run so it always holds exactly the
latest sync's decisions:

MATCH: playlist, track, score, matched Plex ratingKey, and which
candidate field/artist variant won.
NO MATCH: every Plex search hit returned for that title, with its
title/album/grandparentTitle/originalTitle and computed score per
candidate — so a false negative can be diagnosed directly instead of
guessed at.

Co-Authored-By: Claude Sonnet 5 noreply@anthropic.com

Despite the last two rounds of fixes, manual verification is still finding the large majority of "wanted" tracks already own in Plex (e.g. 82+/90 in the last audit). HA's own log rotates/truncates too fast to reliably capture and grep a full sync's worth of matching decisions, which has made root-causing this by log excerpt alone unreliable. _plex_find() now writes a full diagnostic line per track to match_debug.log, created next to engine.py in the installed integration folder (not subject to HA's log rotation or level config), truncated at the start of every match_and_build() run so it always holds exactly the latest sync's decisions: - MATCH: playlist, track, score, matched Plex ratingKey, and which candidate field/artist variant won. - NO MATCH: every Plex search hit returned for that title, with its title/album/grandparentTitle/originalTitle and computed score per candidate — so a false negative can be diagnosed directly instead of guessed at. Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

neerdoc added 1 commit

2026-08-02 21:22:33 +00:00

Write per-track match diagnostics to a file instead of relying on HA's log 59531ba1ac

Despite the last two rounds of fixes, manual verification is still
finding the large majority of "wanted" tracks already own in Plex (e.g.
82+/90 in the last audit). HA's own log rotates/truncates too fast to
reliably capture and grep a full sync's worth of matching decisions,
which has made root-causing this by log excerpt alone unreliable.

_plex_find() now writes a full diagnostic line per track to
match_debug.log, created next to engine.py in the installed integration
folder (not subject to HA's log rotation or level config), truncated at
the start of every match_and_build() run so it always holds exactly the
latest sync's decisions:

- MATCH: playlist, track, score, matched Plex ratingKey, and which
  candidate field/artist variant won.
- NO MATCH: every Plex search hit returned for that title, with its
  title/album/grandparentTitle/originalTitle and computed score per
  candidate — so a false negative can be diagnosed directly instead of
  guessed at.

Co-Authored-By: Claude Sonnet 5 <noreply@anthropic.com>

neerdoc merged commit 3081d7aaca into main

2026-08-02 21:22:36 +00:00

neerdoc referenced this pull request from a commit

2026-08-02 21:22:38 +00:00

Merge pull request 'Write per-track match diagnostics to a file instead of relying on HA's log' (#55) from macsome_14 into main