Files
spacedrive/apps
Jamie PineandClaude Opus 5 ba8efcdda5 Delete the media pipeline that nothing could reach
Deleting locations.triggerJob cut the last live dispatcher for every media
job. What was left was six jobs, six processors, six registered actions and
an 853-line SidecarManager, all keyed on entry rows nothing writes, reachable
from nothing. The UI buttons that fired them were the only thing keeping the
registrations honest, and they were firing into a pipeline that had no input.

The hot tier is unaffected. ThumbService over sources/<id>/thumbs.pvcache is
the live thumbnail path, it is source-scoped already, and it shares no code
with any of this.

Two things survive because they outlive their writers. SidecarKind, variant,
format and status move to domain::sidecar: they are the vocabulary a derived
artifact is addressed by, they are what SdPath::Sidecar parses, and
content_sidecar will be keyed on the same three. And File.sidecars keeps its
shape, empty until something fills it, so the subtitle and thumbstrip lookups
in the client keep compiling.

The EXIF and ffmpeg extraction went with the rest. It read into
image_media_data and video_media_data, which are entity rows on their way
out, and its field mapping is a straight match for the facets declared in
df965b851. When the facet writer gets built it wants rewriting against those
columns rather than resurrecting; git show df965b851~1:core/src/ops/media/
metadata_extractor.rs has the EXIF handling worth copying.

sd-server's sidecar route now builds its path through sd-sidecar-path, the
same way the tauri one already did. Both still serve from <library>/sidecars,
both are equally empty, and L5 re-roots them together.

Co-Authored-By: Claude Opus 5 (1M context) <noreply@anthropic.com>
2026-09-09 20:30:06 -07:00
..
2026-08-25 22:13:17 -07:00