0.6.11: Auto-apply exact-tag format upgrades; self-heal stale pending rows

Two gaps left obvious cases stuck in the manual dedup queue:

1. Format-upgrade pairs found by Pass 2/3 (case-insensitive/normalized
   tag match) still required manual confirm even though the pass
   itself already proved same song via identical tags, and rank_file()
   already guarantees FLAC beats any other format regardless of
   size/dirty-name. A directory-casing difference (e.g. "All About
   This Ep" vs "...EP") meant these never matched the narrower
   same-directory numbered-twin rule from 0.6.10. New is_format_upgrade()
   auto-deletes any Pass 2/3 pair whose extensions differ, gated on
   pass_name so Pass 4 (cross-album fuzzy -- different masters, DJ-mix
   edits, genuinely ambiguous) is untouched and still requires a human.

2. A pending row can go stale without ever being confirmed -- some
   other action (a later auto-apply, upgrade-mp3-to-flac, a manual fix)
   already resolved one side of the pair -- and nothing pruned it, so
   an already-fixed duplicate kept surfacing in the queue indefinitely.
   scan() now sweeps and auto-resolves any pending candidate whose
   keep_path or delete_path no longer exists before persisting new
   ones, so the queue reflects current reality on every run instead of
   accumulating dead entries.

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

README.md

@@ -73,10 +73,10 @@ Optional, add these later if you want them:
 Pull the prebuilt image onto your Docker host:
 
 ```bash
-docker pull git.kretzer.club/andrew/alembic:0.6.6
+docker pull git.kretzer.club/andrew/alembic:0.6.11
 ```
 
-That is the whole install. You do not need to download the source or build anything. The `0.6.6` is the version; you can pin to it so nothing changes under you, or use `latest` to always get the newest.
+That is the whole install. You do not need to download the source or build anything. The `0.6.11` is the version; you can pin to it so nothing changes under you, or use `latest` to always get the newest.
 
 (If you would rather build it yourself from source, you can, but you do not need to.)
 
@@ -136,7 +136,7 @@ Create a file called `docker-compose.yml` on your server (put it wherever you ke
 ```yaml
 services:
   alembic:
-    image: git.kretzer.club/andrew/alembic:0.6.6
+    image: git.kretzer.club/andrew/alembic:0.6.11
     container_name: alembic
     ports:
       - "8420:8420"

app/services/dedup_review_service.py

@@ -27,6 +27,38 @@ def _script_for_pass(pass_name: str) -> str:
     return _FUZZY_SCRIPT if pass_name == _FUZZY_PASS else _SCRIPT
 
 
+def _prune_stale_pending(db) -> int:
+    """Mark pending candidates whose delete_path or keep_path no longer
+    exists as resolved, instead of leaving them to surface forever.
+
+    A pending row can go stale without ever being confirmed through the UI --
+    e.g. a later scan's numbered-twin/format-upgrade auto-apply already
+    removed one side, or upgrade-mp3-to-flac replaced it, or someone deleted
+    it by hand. confirm_and_apply() already does this "already_gone" check
+    for candidates a user explicitly acts on; this generalizes it to run at
+    the start of every scan, so the pending list reflects current reality
+    instead of stale groups that something else already resolved."""
+    now = time.time()
+    rows = db.execute(
+        select(DedupCandidate).where(
+            DedupCandidate.applied == False,  # noqa: E712
+            DedupCandidate.confirmed == False,  # noqa: E712
+            DedupCandidate.ignored == False,  # noqa: E712
+        )
+    ).scalars().all()
+    pruned = 0
+    for c in rows:
+        if not Path(c.delete_path).exists() or not Path(c.keep_path).exists():
+            c.confirmed = True
+            c.confirmed_by = "auto:stale"
+            c.confirmed_at = now
+            c.applied = True
+            pruned += 1
+    if pruned:
+        db.commit()
+    return pruned
+
+
 def _is_pair_ignored(db, path_a: str, path_b: str) -> bool:
     """True if this path pair was ever marked 'keep both', regardless of
     which path was on the keep/delete side that time -- a later scan can
@@ -58,7 +90,9 @@ def _is_pair_already_pending(db, keep_path: str, delete_path: str) -> bool:
     return db.execute(query).first() is not None
 
 
-async def _run_scan(job_key: str, script_name: str, triggered_by: str) -> DedupRun | None:
+async def _run_scan(
+    job_key: str, script_name: str, triggered_by: str, extra_args: tuple[str, ...] = ()
+) -> DedupRun | None:
     """Returns None (persisting nothing) if the job never actually ran --
     e.g. skipped_lock because something else was using the pipeline lock
     at that moment. Same lesson as genre_review_service.run(): recording a
@@ -67,7 +101,7 @@ async def _run_scan(job_key: str, script_name: str, triggered_by: str) -> DedupR
     pending-candidates list isn't scoped to "latest run only"."""
     script = str(settings.pipeline_dir / "lib" / script_name)
     job_run, output = await pipeline_runner.run_job_capture(
-        job_key, [script, "--json"], triggered_by=triggered_by, timeout=_JOB_TIMEOUT_SECONDS
+        job_key, [script, "--json", *extra_args], triggered_by=triggered_by, timeout=_JOB_TIMEOUT_SECONDS
     )
     if job_run.status != "success":
         return None
@@ -76,6 +110,8 @@ async def _run_scan(job_key: str, script_name: str, triggered_by: str) -> DedupR
 
     db = SessionLocal()
     try:
+        _prune_stale_pending(db)
+        now = time.time()
         dedup_run = DedupRun(
             started_at=job_run.started_at,
             finished_at=job_run.finished_at,
@@ -91,7 +127,30 @@ async def _run_scan(job_key: str, script_name: str, triggered_by: str) -> DedupR
 
         skipped_ignored = 0
         skipped_duplicate = 0
+        auto_applied = 0
         for c in candidates:
+            if c.get("auto_applied"):
+                # dedup-library.sh already deleted this pair unconditionally
+                # (identical numbered-sibling twin: same dir/name/ext, no
+                # tag/fuzzy ambiguity involved) -- record it pre-applied for
+                # audit history; it never shows up as a pending review item.
+                db.add(
+                    DedupCandidate(
+                        dedup_run_id=dedup_run.id,
+                        pass_name=c.get("pass", "unknown"),
+                        keep_path=c["keep_path"],
+                        delete_path=c["delete_path"],
+                        delete_id=c.get("delete_id"),
+                        delete_size_bytes=c.get("delete_size_bytes"),
+                        confirmed=True,
+                        confirmed_by="auto:numbered_twin",
+                        confirmed_at=now,
+                        applied=True,
+                    )
+                )
+                auto_applied += 1
+                db.flush()
+                continue
             if _is_pair_ignored(db, c["keep_path"], c["delete_path"]):
                 skipped_ignored += 1
                 continue
@@ -116,8 +175,11 @@ async def _run_scan(job_key: str, script_name: str, triggered_by: str) -> DedupR
         db.commit()
         db.refresh(dedup_run)
         skipped_total = skipped_ignored + skipped_duplicate
+        if skipped_total or auto_applied:
             if skipped_total:
                 dedup_run.kept = (dedup_run.kept or 0) + skipped_total
+            if auto_applied:
+                dedup_run.deleted = auto_applied
             db.commit()
             db.refresh(dedup_run)
         return dedup_run
@@ -126,12 +188,23 @@ async def _run_scan(job_key: str, script_name: str, triggered_by: str) -> DedupR
 
 
 async def scan(triggered_by: str = "manual") -> DedupRun | None:
-    """Dry-run dedup-library.sh --json, persist every candidate deletion
-    into a fresh dedup_runs/dedup_candidates pair. Never deletes anything
-    -- the scheduled maintenance:dedup job also only ever calls this (no
-    --apply), matching the false-negative-biased dedup preference; actual
-    deletion only ever happens through confirm_and_apply() below."""
-    return await _run_scan("dedup:scan", _SCRIPT, triggered_by)
+    """Dry-run dedup-library.sh --json (plus --auto-apply-numbered), persist
+    every candidate deletion into a fresh dedup_runs/dedup_candidates pair.
+
+    Two kinds of match need no human judgment and get deleted unconditionally
+    by dedup-library.sh itself (see is_numbered_twin/is_format_upgrade there):
+    identical numbered-sibling twins (same dir, same name, same extension,
+    differing only by the ".N" collision suffix), and format upgrades within
+    an EXACT tag match from Pass 2/3 (case-insensitive/normalized already
+    proved same song via identical tags, so a differing extension there is
+    just "better format vs. worse"). Everything else -- including Pass 4's
+    cross-album fuzzy matches, where a wrong auto-delete could take out a
+    genuinely different version (a different album pressing, a DJ-mix edit,
+    etc.) -- stays a dry-run candidate awaiting manual confirm_and_apply()
+    below; that's the false-negative-biased preference for anything with real
+    ambiguity. Auto-applied deletions are reported here already applied,
+    purely for audit visibility."""
+    return await _run_scan("dedup:scan", _SCRIPT, triggered_by, extra_args=("--auto-apply-numbered",))
 
 
 async def scan_fuzzy(triggered_by: str = "manual") -> DedupRun | None:

pipeline/bin/run-playlist.sh

@@ -96,6 +96,21 @@ else
   log "Fetched $TRACK_COUNT track(s) from Spotify"
 fi
 
+# ==== Seed the pinned sldl skip index if it's missing or stranded ====
+# The conf pins index-path to $DROPBOX/_index.csv (see _template.conf: sldl's
+# default index location is derived from the input, so an input change strands
+# the old index and the whole playlist re-downloads -- that's what produced
+# the 2026-07-16 duplicate flood). If the pinned file is missing, or an index
+# in one of sldl's old input-named subfolders is newer (i.e. sldl last wrote
+# somewhere else), fold them all into the pinned location first.
+INDEX_FILE="$DROPBOX/_index.csv"
+if [[ ! -s "$INDEX_FILE" ]] || \
+   [[ -n "$(find "$DROPBOX" -mindepth 2 -maxdepth 2 -name _index.csv -newer "$INDEX_FILE" -print -quit 2>/dev/null)" ]]; then
+  log "Seeding pinned sldl index at $INDEX_FILE from prior indexes"
+  python3 "${PIPELINE_DIR:-/app/pipeline}/lib/merge-sldl-indexes.py" "$DROPBOX" "$INDEX_FILE" >> "$LOG" 2>&1 \
+    || log "WARNING: index merge failed -- sldl may re-download tracks the library already has"
+fi
+
 # ==== Run sldl (vendored binary, subprocess of this container) ====
 log "Running sldl for: $PLAYLIST_NAME"
 

pipeline/configs/_template.conf

@@ -31,6 +31,18 @@ path = SLDL_DROPBOX_ROOT/PLAYLIST_NAME
 playlist-path = SLDL_DROPBOX_ROOT/PLAYLIST_NAME/_sldl.m3u8
 write-playlist = true
 
+# ==== Skip index (pinned!) ====
+# sldl's already-downloaded index defaults to {path}/{playlist-name}/_index.csv
+# where {playlist-name} is derived from the INPUT: the Spotify playlist's
+# display name for spotify input, the CSV filename for csv input. beets moves
+# every download out of the dropbox, so this index is the ONLY thing standing
+# between a nightly run and re-downloading the whole playlist. When 0.6.2
+# switched input from Spotify URL to CSV, the derived name changed, sldl
+# started a fresh empty index in a new subfolder, and every playlist
+# re-downloaded in full on 2026-07-16. Pinning the path here decouples the
+# index from input naming so that can never happen again.
+index-path = SLDL_DROPBOX_ROOT/PLAYLIST_NAME/_index.csv
+
 # ==== Naming ====
 name-format = {artist} - {title}
 

pipeline/lib/dedup-library.sh

@@ -34,11 +34,26 @@ set -euo pipefail
 APPLY=0
 JSON_MODE=0
 ONLY_PATHS_FILE=""
+AUTO_APPLY_NUMBERED=0
 while [[ $# -gt 0 ]]; do
   case "$1" in
     --apply) APPLY=1; shift ;;
     --json) JSON_MODE=1; shift ;;
     --only-paths) ONLY_PATHS_FILE="$2"; shift 2 ;;
+    # Delete two kinds of match unconditionally, independent of
+    # --apply/--only-paths, while everything else stays dry-run/log-only:
+    #   - numbered-sibling twins (see is_numbered_twin): same dir, same name,
+    #     same extension, just the ".N" collision suffix -- the same download
+    #     landing twice.
+    #   - format upgrades within an exact tag match (see is_format_upgrade):
+    #     Pass 2/3 already proved same song via identical tags; a different
+    #     extension there just means "better format vs. worse", never a
+    #     different version.
+    # Excludes Pass 4 (cross-album fuzzy) entirely -- matching across
+    # different albums/directories is exactly where a different master or
+    # DJ-mix edit can share tags without being interchangeable, so it always
+    # needs a human. See dedup_review_service.scan().
+    --auto-apply-numbered) AUTO_APPLY_NUMBERED=1; shift ;;
     *) shift ;;
   esac
 done
@@ -78,11 +93,63 @@ fi
 # Counters: tracked via log grep at the end (avoids bash subshell pitfalls).
 # Functions write KEEP/DELETE lines with consistent prefixes; summary greps them.
 
-# Convert container path (/music/...) to host path (${MUSIC_DATA_DIR:-/data/music}/Library/...)
-host_path() { echo "${MUSIC_DATA_DIR:-/data/music}/Library${1#/music}"; }
+# Resolve a beets-displayed path to a real path in this container. Since the
+# 2026-07-08 path rewrite, beets displays ${MUSIC_DATA_DIR:-/data/music}/Library/...
+# paths that are directly usable — pass those through UNCHANGED. Only legacy
+# /music/... display paths (the pre-rewrite transitional mount) still need the
+# prefix swap. Blindly prepending the library dir to an already-correct path
+# (the old behavior) produced /data/music/Library/data/music/Library/... —
+# nonexistent, so every candidate failed the -f check and every pass reported
+# 0 groups from 2026-07-08 until this fix.
+host_path() {
+  local p="$1"
+  if [[ "$p" == /music/* ]]; then
+    echo "${MUSIC_DATA_DIR:-/data/music}/Library${p#/music}"
+  else
+    echo "$p"
+  fi
+}
 
 SEP=$'\x1c'  # ASCII file separator — safe with any music metadata
 
+# True if two paths are the SAME directory, SAME extension, and identical
+# basenames except for a numeric ".N" infix sldl/beets inserts on a filename
+# collision (e.g. "Title.flac" vs "Title.1.flac"). This is deliberately
+# narrower than any tag-based pass: no fuzzy matching, no cross-album
+# ambiguity, just the literal same download landing twice. Used to
+# auto-delete regardless of which pass found the pair -- Pass 1 only fires
+# when the canonical name is a beets ghost; the very common case where BOTH
+# copies are already tracked in beets (so Pass 1 defers) surfaces instead via
+# Pass 2/3's tag-based grouping, and still deserves the same free pass.
+is_numbered_twin() {
+  local a="$1" b="$2"
+  [[ "${a%/*}" == "${b%/*}" ]] || return 1
+  local ba="${a##*/}" bb="${b##*/}"
+  local ext="${ba##*.}"
+  [[ "${ext,,}" == "${bb##*.}" ]] || return 1
+  [[ "$(echo "$ba" | sed -E "s/\.[0-9]+\.${ext}\$/.${ext}/")" \
+     == "$(echo "$bb" | sed -E "s/\.[0-9]+\.${ext}\$/.${ext}/")" ]]
+}
+
+# True if this pair is a same-song format upgrade found by an EXACT tag match
+# (Pass 2 case-insensitive or Pass 3 normalized -- both require identical
+# albumartist+album+title after case/punctuation folding, so "same song" is
+# already proven by the pass itself) where the two files simply have
+# different extensions. rank_file() always ranks FLAC ahead of any other
+# format regardless of size/dirty-name/etc, so within a tag-exact group a
+# cross-extension pair unconditionally means "the format-preferred copy vs.
+# a lesser one" -- no judgment call. Deliberately excludes Pass 4
+# (cross_album_fuzzy): that pass matches by artist+title-base ACROSS
+# different albums/directories, which is exactly where a different master,
+# DJ-mix edit, or reissue can share tags without being an interchangeable
+# copy -- those still need a human to look at album/context before deleting
+# either side.
+is_format_upgrade() {
+  local pass="$1" keep="$2" delete="$3"
+  [[ "$pass" == "case_insensitive" || "$pass" == "normalized" ]] || return 1
+  [[ "${keep##*.}" != "${delete##*.}" ]]
+}
+
 # Rank a file: lower = better. FLAC > MP3 > other (WAV/etc.); clean filename
 # > sync-conflict artifact (-DESKTOP-, (1), .1.); extended/club mix > radio
 # edit at same format; then larger wins.
@@ -163,7 +230,7 @@ process_group() {
   esac
 
   local first=1
-  local keep_path=""
+  local keep_path="" keep_id="" any_auto=0
   while IFS="$SEP" read -r _sk id hp; do
     [[ -z "$hp" ]] && continue
     local size_bytes; size_bytes=$(stat -c '%s' "$hp" 2>/dev/null || echo 0)
@@ -171,14 +238,31 @@ process_group() {
     if [[ $first -eq 1 ]]; then
       log "  KEEP   ($size_h) $hp"
       keep_path="$hp"
+      keep_id="$id"
       first=0
     else
       log "  DELETE ($size_h) $hp"
-      json_emit "{\"pass\":\"${pass_name}\",\"keep_path\":\"$(json_escape "$keep_path")\",\"delete_path\":\"$(json_escape "$hp")\",\"delete_size_bytes\":${size_bytes}}"
+
+      # Auto-apply-numbered deletes numbered_sibling matches unconditionally,
+      # bypassing --only-paths (that gate exists only for the manual confirm
+      # flow, which never runs alongside this flag). All other passes stay
+      # dry-run/log-only here regardless.
+      local do_delete=0 auto_applied=0
       if [[ $APPLY -eq 1 ]]; then
+        do_delete=1
+      elif [[ $AUTO_APPLY_NUMBERED -eq 1 ]] \
+        && { is_numbered_twin "$keep_path" "$hp" || is_format_upgrade "$pass_name" "$keep_path" "$hp"; }; then
+        do_delete=1
+        auto_applied=1
+        any_auto=1
+      fi
+      local auto_json="false"; [[ $auto_applied -eq 1 ]] && auto_json="true"
+      json_emit "{\"pass\":\"${pass_name}\",\"keep_path\":\"$(json_escape "$keep_path")\",\"delete_path\":\"$(json_escape "$hp")\",\"delete_size_bytes\":${size_bytes},\"auto_applied\":${auto_json}}"
+
+      if [[ $do_delete -eq 1 ]]; then
         # With --only-paths given, only delete entries the caller explicitly
         # confirmed; without it, delete everything (original behavior).
-        if [[ -n "$ONLY_PATHS_FILE" && -z "${ONLY_PATHS[$hp]:-}" ]]; then
+        if [[ $auto_applied -eq 0 && -n "$ONLY_PATHS_FILE" && -z "${ONLY_PATHS[$hp]:-}" ]]; then
           log "    (skipped — not in --only-paths confirm list)"
         elif [[ -n "$id" ]]; then
           # In beets — beet remove -d removes from DB + disk. Best-effort: a
@@ -192,6 +276,18 @@ process_group() {
       fi
     fi
   done < <(echo "$ranked" | grep -v '^$' | sort -n)
+
+  # An auto-applied numbered-twin deletion can leave the numbered-named file
+  # as the sole survivor (it won on size despite the dirty-filename penalty).
+  # Rename it back to the canonical name so the library doesn't accumulate
+  # ".1."/".2." filenames for tracks that no longer have a duplicate. Only
+  # auto_applied triggers this (never a manual --only-paths confirm, which
+  # may leave the sibling undeleted if the user didn't confirm it -- renaming
+  # then would collide); only when it's actually in beets (a ghost survivor
+  # has nothing to move); only when the name still has the dirty infix.
+  if [[ $any_auto -eq 1 && -n "$keep_id" && "$keep_path" =~ \.[0-9]+\.[A-Za-z0-9]+$ ]]; then
+    beet move "id:${keep_id}" >> "$LOG" 2>&1 || log "    WARN: beet move id:${keep_id} failed (rename survivor)"
+  fi
   return 0  # explicit: the loop's last command status must not leak out under set -e
 }
 
@@ -202,9 +298,10 @@ process_group() {
 log ""
 log "[$(date -Iseconds)] === Pass 1: numbered siblings ==="
 
-# Dump id + container path from beets once. beets normalizes $path for
-# display (always /music/-prefixed) even though the DB stores a mix of
-# absolute and relative — so the display paths are safe to compare/convert,
+# Dump id + display path from beets once. beets normalizes $path for display
+# (real /data/music/Library/... paths since the 2026-07-08 rewrite; /music/...
+# on any legacy entry) even though the DB stores a mix of absolute and
+# relative — so the display paths are safe to compare/convert via host_path(),
 # and the ids are what we hand to beet remove/move.
 beet ls -f "\$id${SEP}\$path" 2>/dev/null > /tmp/beets-id-paths.txt
 cut -d"$SEP" -f2- /tmp/beets-id-paths.txt > /tmp/beets-all-paths.txt
@@ -231,18 +328,22 @@ while IFS="$SEP" read -r numbered_id numbered_cp; do
 
   if [[ $score_canonical -le $score_numbered ]]; then
     # Canonical wins: delete numbered (in beets), keep canonical (ghost)
-    process_group "P1 numbered-sibling: ${numbered_cp##/music/}" \
+    process_group "P1 numbered-sibling: ${numbered_cp##*/Library/}" \
       "$host_canonical" "${numbered_id}${SEP}${numbered_cp}"
-    if [[ $APPLY -eq 1 ]]; then
+    if [[ $APPLY -eq 1 || $AUTO_APPLY_NUMBERED -eq 1 ]]; then
       # canonical is now a ghost file; import it into beets
       beet import -q -s "${canonical_cp}" >> "$LOG" 2>&1 || log "  WARN: beet import failed for ${canonical_cp}"
     fi
   else
     # Numbered wins (canonical is ghost): rm the ghost, then beet move to rename
     # the numbered file to the canonical path (id query — see process_group).
-    process_group "P1 numbered-sibling: ${numbered_cp##/music/}" \
+    # Safe to reach with just AUTO_APPLY_NUMBERED: this loop only ever builds
+    # genuine numbered-sibling pairs (see is_numbered_twin), and process_group
+    # above already deleted the ghost canonical before this runs, so the move
+    # target is guaranteed clear -- no rename collision.
+    process_group "P1 numbered-sibling: ${numbered_cp##*/Library/}" \
       "${numbered_id}${SEP}${numbered_cp}" "$host_canonical"
-    if [[ $APPLY -eq 1 ]]; then
+    if [[ $APPLY -eq 1 || $AUTO_APPLY_NUMBERED -eq 1 ]]; then
       beet move "id:${numbered_id}" >> "$LOG" 2>&1 || log "  WARN: beet move id:${numbered_id} failed"
     fi
   fi

pipeline/lib/export-laptop-playlists.py

@@ -57,11 +57,12 @@ OUT_DIR = f"{_MUSIC_DATA_DIR}/playlists-laptop"
 # Absolute Windows path to the laptop's library root.
 LIBRARY_LAPTOP_ROOT = r"C:\Music\Library"
 
-# Beets stores paths with this prefix (the container-side mount of the
-# library); strip it to get the path relative to the library root. This is
-# still "/music/" through the migration's Stage 0-3 transitional mount;
-# revisit at Stage 4 once beets' directory: becomes MUSIC_DATA_DIR/Library.
-BEETS_LIBRARY_PREFIX = "/music/"
+# Prefixes beets may display library paths under, tried in order: the real
+# library dir (everything since the 2026-07-08 path rewrite) and the legacy
+# transitional mount (any stale pre-rewrite entry). Strip whichever matches
+# to get the path relative to the library root; when neither matches the
+# track is outside the library and is skipped.
+BEETS_LIBRARY_PREFIXES = (f"{_MUSIC_DATA_DIR}/Library/", "/music/")
 
 M3U_EXT = ".m3u"
 # Older formats from earlier iterations of this script — swept by reconcile.
@@ -157,9 +158,12 @@ def build_beets_index() -> dict[tuple[str, str], list[tuple[str, str, str, str]]
         if len(parts) != 4:
             continue
         aa, alb, ti, path = parts
-        if not path.startswith(BEETS_LIBRARY_PREFIX):
+        rel = next(
+            (path[len(p):] for p in BEETS_LIBRARY_PREFIXES if path.startswith(p)),
+            None,
+        )
+        if rel is None:
             continue
-        rel = path[len(BEETS_LIBRARY_PREFIX):]
         idx.setdefault((_normkey(alb), _normkey(ti)), []).append((aa, alb, ti, rel))
     return idx
 

pipeline/lib/fix-track-metadata.py

@@ -49,8 +49,10 @@ _ALEMBIC_CONFIG_DIR = os.environ.get("ALEMBIC_CONFIG_DIR", "/config")
 _MUSIC_DATA_DIR = os.environ.get("MUSIC_DATA_DIR", "/data/music")
 
 LIBRARY = f"{_MUSIC_DATA_DIR}/Library"
-# Still "/music" through the migration's Stage 0-3 transitional beets mount;
-# revisit at Stage 4 once beets' directory: becomes MUSIC_DATA_DIR/Library.
+# Legacy prefix from the migration's transitional beets mount. Since the
+# 2026-07-08 path rewrite beets displays real LIBRARY paths, so this only
+# matters for accepting pasted /music/... paths as input and as a fallback
+# beets query form for any stale pre-rewrite DB entry.
 CONTAINER_PREFIX = "/music"
 SPOTIFY_ENV = f"{_ALEMBIC_CONFIG_DIR}/pipeline/_spotify.env"
 SPOTIFY_GENRE = f"{os.environ.get('PIPELINE_DIR', '/app/pipeline')}/lib/spotify-genre.py"
@@ -316,7 +318,11 @@ def resolve_input_path(arg):
 
 
 def beets_id_for(host_path):
-    cp = host_to_container(host_path)
+    # Real path first (how beets displays everything since the 2026-07-08
+    # path rewrite), legacy /music/... form second for any stale entry.
+    # Querying only the legacy form (the old behavior) matched nothing after
+    # the rewrite, so retag-from-url silently skipped beet update/move.
+    for cp in (host_path, host_to_container(host_path)):
         r = subprocess.run(
             ["beet", "ls", "-f", "$id|||$path", f"path:{cp}"],
             capture_output=True, text=True, check=True)

pipeline/lib/merge-sldl-indexes.py

@@ -0,0 +1,121 @@
+#!/usr/bin/env python3
+"""Fold every sldl skip index under a playlist's dropbox into one pinned file.
+
+sldl names its per-playlist index folder after the *input*: the Spotify
+playlist's display name for spotify input, the CSV filename stem for csv
+input. So when 0.6.2 switched input from Spotify URL to CSV, sldl started a
+fresh empty index in a new subfolder, saw no download history, and
+re-downloaded every playlist in full (2026-07-16). The rendered confs now pin
+index-path to <dropbox>/<playlist>/_index.csv; this script seeds that pinned
+file from all the indexes sldl left behind (run-playlist.sh calls it whenever
+the pinned file is missing or older than a stranded one).
+
+Usage: merge-sldl-indexes.py <playlist_dropbox_dir> <output_index_csv>
+
+Merge rules, per (artist, title) lowercased:
+  - the row from the newest index file wins;
+  - EXCEPT when that row is a failure (state 2) and any older index has the
+    track as downloaded (state 1 or 3): then the newest row's identity
+    (artist/album/title/length -- matching what the current input will
+    present; Spotify length rounding drifted between the old extractor and
+    our CSV) is kept but marked state 3 (already downloaded), so sldl does
+    not re-fetch a track the library already holds;
+  - rows only present in older indexes are kept as-is (tracks since removed
+    from the playlist; harmless, and they keep their history if re-added).
+"""
+
+import csv
+import sys
+from pathlib import Path
+
+HEADER = ["filepath", "artist", "album", "title", "length", "tracktype", "state", "failurereason"]
+DOWNLOADED_STATES = {"1", "3"}  # 1 = downloaded this run, 3 = found in index previously
+FAILED_STATE = "2"
+
+
+def read_rows(path: Path) -> list[dict]:
+    rows = []
+    with open(path, newline="", encoding="utf-8") as fh:
+        reader = csv.DictReader(fh)
+        for row in reader:
+            if row.get("artist") is None or row.get("title") is None:
+                continue
+            rows.append({k: (row.get(k) or "") for k in HEADER})
+    return rows
+
+
+def norm_key(row: dict) -> tuple:
+    # Primary artist only: sldl's old Spotify extractor recorded just the
+    # first artist, while spotify-playlist-csv.py joins all of them with
+    # ", " -- keying on the full string would miss every multi-artist track
+    # when recovering history across the two index generations. First
+    # comma-segment matches both forms (and both sides of a comma-in-name
+    # artist like "Tyler, The Creator" truncate identically).
+    return (row["artist"].split(",")[0].strip().lower(), row["title"].strip().lower())
+
+
+def main() -> int:
+    if len(sys.argv) != 3:
+        print(__doc__, file=sys.stderr)
+        return 2
+
+    dropbox = Path(sys.argv[1])
+    out_path = Path(sys.argv[2])
+    if not dropbox.is_dir():
+        print(f"[merge-sldl-indexes] not a directory: {dropbox}", file=sys.stderr)
+        return 2
+
+    # Every index at the dropbox root or one level down (sldl's input-named
+    # subfolders), including the pinned output itself if it already exists --
+    # newest first, so the most recent record of each track wins.
+    candidates = sorted(
+        set(dropbox.glob("_index.csv")) | set(dropbox.glob("*/_index.csv")),
+        key=lambda p: p.stat().st_mtime,
+        reverse=True,
+    )
+    if not candidates:
+        print(f"[merge-sldl-indexes] no _index.csv found under {dropbox}; nothing to seed")
+        return 0
+
+    merged: dict[tuple, dict] = {}
+    recovered = 0
+    for path in candidates:
+        try:
+            rows = read_rows(path)
+        except (OSError, csv.Error) as exc:
+            print(f"[merge-sldl-indexes] skipping unreadable {path}: {exc}", file=sys.stderr)
+            continue
+        print(f"[merge-sldl-indexes] {path}: {len(rows)} rows")
+        for row in rows:
+            key = norm_key(row)
+            kept = merged.get(key)
+            if kept is None:
+                merged[key] = row
+            elif kept["state"] == FAILED_STATE and row["state"] in DOWNLOADED_STATES:
+                # Newest attempt failed but an older index proves we already
+                # have this track: keep the newest identity fields, take the
+                # old filepath (informational only; skip-mode index never
+                # checks the file on disk), and mark it downloaded.
+                kept["filepath"] = row["filepath"]
+                kept["state"] = "3"
+                kept["failurereason"] = "0"
+                recovered += 1
+
+    out_path.parent.mkdir(parents=True, exist_ok=True)
+    tmp = out_path.with_suffix(".csv.tmp")
+    with open(tmp, "w", newline="", encoding="utf-8") as fh:
+        writer = csv.DictWriter(fh, fieldnames=HEADER)
+        writer.writeheader()
+        writer.writerows(merged.values())
+    tmp.replace(out_path)
+
+    downloaded = sum(1 for r in merged.values() if r["state"] in DOWNLOADED_STATES)
+    print(
+        f"[merge-sldl-indexes] wrote {out_path}: {len(merged)} tracks "
+        f"({downloaded} downloaded, {recovered} recovered from older indexes)"
+    )
+    return 0
+
+
+if __name__ == "__main__":
+    sys.exit(main())

pipeline/lib/pipeline-status.sh

@@ -18,8 +18,10 @@
 # swallow the whole message.
 #
 # Reports on:
-#   - Sibling service reachability (slskd, navidrome) via HTTP, not docker ps —
-#     alembic has no Docker socket access
+#   - Sibling service reachability (navidrome) via HTTP, not docker ps —
+#     alembic has no Docker socket access. slskd is deliberately NOT checked:
+#     it is not part of the pipeline (sldl is its own Soulseek client) — it
+#     runs on the host purely as the user's own file-sharing presence.
 #   - Today's runs: playlist syncs, Bandcamp sync, manual imports, dedup
 #   - Weekly maintenance freshness: strip-mb-tags, strip-watermark-art,
 #     scrub-watermark-text, clean-sldl-index, spotify-genre
@@ -188,7 +190,6 @@ dedup_today_status() {
       mark_warn "$name" "unreachable at $url"
     fi
   }
-  check_http slskd     "http://gluetun:5030/"
   check_http navidrome "http://navidrome:4533/rest/ping.view"
 
   # 2. Today's runs

pipeline/lib/replace-with-better.sh

@@ -195,7 +195,18 @@ while IFS= read -r -d '' new_file; do
   existing_id="${match%%$'\x1f'*}"
   existing_container="${match#*$'\x1f'}"
 
+  # beets displays real ${MUSIC_DATA_DIR}/Library/... paths since the
+  # 2026-07-08 path rewrite -- use them as-is; only a legacy /music/...
+  # display path still needs the prefix swap. Blindly prepending (the old
+  # behavior) doubled the prefix, so every match logged [stale-db], the MP3
+  # never got replaced, and the leftover-import step at the bottom imported
+  # the staged FLAC as a NEW track -- producing exactly the flac+mp3 twins
+  # this script exists to prevent.
+  if [[ "$existing_container" == /music/* ]]; then
     existing="${MUSIC_DATA_DIR:-/data/music}/Library${existing_container#/music}"
+  else
+    existing="$existing_container"
+  fi
   if [[ ! -f "$existing" ]]; then
     echo "[stale-db]       beets has $existing_container but file missing" | tee -a "$LOG"
     continue