Add defensible orphaned audio detection

ROADMAP.md

@@ -19,21 +19,21 @@
 
 - [x] `serato-doctor analyze` command
 - [x] Duplicate filename detection
-- [ ] Duplicate audio hash detection
+- [x] Duplicate audio hash detection
 - [x] Broken symlink detection
-- [ ] Orphaned audio detection
+- [x] Orphaned audio detection
 - [ ] OneDrive rename detection
 - [x] Crate classification: static vs smart/dynamic
 - [x] Library health score
 
 ## v0.3 — Safe Repair
 
-- [ ] Dry-run repair plan
+- [x] Dry-run repair plan
-- [ ] Backup before repair
+- [x] Backup before repair
-- [ ] Compatibility symlink creation
+- [x] Compatibility symlink creation
 - [ ] Compatibility copy creation
 - [ ] Rename repair
-- [ ] Rollback log
+- [x] Rollback log and recovery center
 
 ## v0.4 — Migration Wizard
 

docs/design/audio-comparison.md

@@ -0,0 +1,32 @@
+# Audio comparison for duplicate review
+
+## Problem
+
+Filenames alone are weak evidence. DJs need to hear each candidate and locate it
+on disk before deciding which copy is authoritative.
+
+## Design
+
+Exact-duplicate and cloud-conflict groups expose a shared audio player with a
+preview action for each file. Switching files reuses the same player, making
+back-to-back comparison quick. A separate action reveals the selected file in
+macOS Finder.
+
+The browser never receives unrestricted filesystem access. Each analyzed audio
+path gets a short, process-local HMAC token. Preview and Finder endpoints reject
+altered, expired, missing, non-audio, or otherwise unsigned paths. Audio serving
+supports HTTP byte ranges so playback can seek without loading an entire track.
+
+## Edge cases
+
+- A file moved after analysis is rejected.
+- Forged or stale tokens cannot select another local file.
+- Preview controls remain useful if autoplay is blocked because native audio
+  controls stay visible.
+- Finder failures are reported beside the button without affecting analysis.
+
+## Tests
+
+- Valid signed audio resolves to the analyzed file.
+- A modified token is rejected.
+- Duplicate detail payloads include preview and Finder controls for every file.

docs/design/backup-recovery-center.md

@@ -0,0 +1,30 @@
+# Backup recovery center
+
+## Problem
+
+An immediate undo button is not enough. A DJ may restart Serato Doctor, notice
+an issue later, or need to understand how much disk space safety snapshots use.
+
+## Architecture
+
+The recovery center reads the existing JSON manifests beneath
+`_Serato_/.serato-doctor-backups/`. It reports creation time, selected keeper,
+affected paths, restore status, and snapshot size. Invalid or incomplete backup
+folders are ignored rather than presented as restorable.
+
+Restore remains conservative: it only replaces a symbolic-link alias and
+refuses to overwrite a real file. A successful restore records its timestamp in
+the manifest so the UI cannot accidentally offer the same rollback twice.
+
+## Edge cases
+
+- No backup directory is treated as an empty history.
+- Malformed manifests do not prevent valid backups from loading.
+- Restored snapshots remain visible for audit purposes and disk accounting.
+- Backups can only be restored through the Serato library that owns them.
+
+## Tests
+
+- Backup history reports size and ready status.
+- Restore changes the persistent status to restored.
+- Static recovery assets are included in the package.

docs/design/batch-duplicate-review.md

@@ -0,0 +1,37 @@
+# Batch duplicate review
+
+## Problem
+
+Opening a diagnostic, opening each group again, scrolling to a separate repair
+form, and applying one change at a time creates unnecessary friction and many
+small backups.
+
+## Design
+
+Duplicate and cloud-conflict diagnostics now open as a sequential review queue.
+Each group presents its candidates together with audio preview and Finder
+controls. Choosing a keeper records the decision and advances immediately.
+Uncertain groups can be skipped without changing them.
+
+All approved decisions are submitted as one dry-run plan. Apply creates one
+backup containing every affected audio file plus the Serato metadata snapshot,
+then installs all compatibility shortcuts as one rollback unit. Any validation
+or filesystem failure prevents or rolls back the batch.
+
+The dry-run opens in a closable modal that names every selected keeper and every
+path that will become a compatibility shortcut. Closing the modal preserves the
+queue position and all decisions so the DJ can continue reviewing before apply.
+
+## Edge cases
+
+- The same group cannot be submitted twice.
+- A file cannot be both a keeper and a disposable file across choices.
+- Every group is rescanned and revalidated before preview and apply.
+- Skipped and unreviewed groups remain untouched.
+- Revisiting a group preserves and visibly marks its selected keeper.
+
+## Tests
+
+- Two groups produce one backup and one restorable receipt.
+- Batch preview reports combined changes without modifying files.
+- Existing single-group repair remains a wrapper around the batch engine.

docs/design/diagnostic-drilldowns.md

@@ -0,0 +1,35 @@
+# Diagnostic Drill-Downs
+
+## Problem
+
+Aggregate health numbers are useful, but DJs need to understand what is behind
+each number before they can trust it. A count like "60 missing tracks" should be
+clickable enough to answer: which tracks, which saved paths, and why did Serato
+Doctor count them?
+
+## Architecture
+
+The health engine remains responsible for aggregate scoring. The web layer adds
+a separate `details` payload beside the existing health report so the UI can show
+examples without changing the core score model.
+
+The drill-down payload is intentionally capped. Serato Doctor should explain the
+finding quickly in the local browser, not dump an entire user library into the
+page.
+
+## Edge Cases
+
+- Smart/dynamic crate references stay excluded from old-reference scoring.
+- The same missing filename can appear more than once in Serato's database.
+- Duplicate filename and cloud-conflict groups are informational only.
+- Suggested matches remain read-only evidence and must never trigger repair.
+- Empty drill-downs should feel reassuring, not broken.
+
+## Testing
+
+- Web analysis should include detail sections for old crate references,
+  suggested matches, unused tracks, duplicates, cloud conflicts, broken
+  symlinks, and missing Serato database tracks.
+- Static assets should expose clickable diagnostic hooks.
+- Browser behavior should be verified manually when the local browser policy
+  allows access to the development server.

docs/design/duplicate-audio-hashing.md

@@ -0,0 +1,34 @@
+# Duplicate audio hashing
+
+## Problem
+
+Matching filenames do not prove matching content. Two files can be different DJ
+edits, masters, encodes, or entirely different tracks. Removing either without
+stronger evidence is unsafe.
+
+## Design
+
+Serato Doctor computes SHA-256 only for the duplicate group currently being
+reviewed. This avoids hashing an entire 22,000-track library during every scan.
+The review queue labels a group as byte-for-byte identical or warns that its
+files differ. File size and a short fingerprint remain available as supporting
+detail.
+
+The final batch dry-run recomputes every approved comparison server-side and
+includes the status beside each keeper decision. The hash is evidence, not an
+automatic repair decision; the DJ remains in control.
+
+## Edge cases
+
+- Identical audio stored under different filenames is recognized.
+- Metadata changes inside an audio container produce a different byte hash and
+  therefore a conservative warning.
+- Missing or unreadable files fail verification rather than being treated as
+  identical.
+- Hashing is streamed in chunks rather than loading whole tracks into memory.
+
+## Tests
+
+- Equal byte content produces identical SHA-256 fingerprints.
+- Different content is labeled different.
+- Batch previews carry the server-computed confidence status.

docs/design/duplicate-repair.md

@@ -0,0 +1,39 @@
+# Duplicate repair
+
+## Problem
+
+Duplicate and cloud-conflict files waste space, but deleting either path can
+unmap tracks in crates or Serato's database. DJs need to choose the authoritative
+copy and understand every change before it happens.
+
+## Design
+
+The web interface requires an analysis, an explicit keeper selection, and a
+dry-run preview. Applying the plan first creates a timestamped backup beneath
+`_Serato_/.serato-doctor-backups/`. The snapshot contains every replaced audio
+file, loaded crate/smart-crate metadata, database V2, and a JSON restore manifest.
+
+The non-kept audio path is then replaced with a symbolic link to the keeper.
+This removes the extra audio payload while preserving every existing saved path.
+Crate files and database V2 are never rewritten. The UI exposes immediate
+restore using the manifest.
+
+Users may retain all backups or set a positive rotation limit. Rotation occurs
+only after a repair succeeds.
+
+## Edge cases
+
+- The duplicate group is rescanned and validated immediately before preview and
+  apply.
+- Existing symlinks cannot be selected as disposable duplicate files.
+- A partial failure restores already-changed files before reporting the error.
+- Restore refuses to overwrite a real file.
+- Healthy symlink aliases are excluded from future duplicate counts.
+
+## Tests
+
+- Backup creation includes audio and Serato metadata.
+- The old path resolves to the selected keeper after repair.
+- Restore returns the original file contents.
+- Invalid keeper choices are rejected.
+- Limited and unlimited retention behave deterministically.

docs/design/orphaned-audio.md

@@ -0,0 +1,34 @@
+# Orphaned audio detection
+
+## Problem
+
+“Not in a crate” does not mean unused. Many DJs keep active tracks in Serato's
+main library without placing them in a crate. Treating those files as orphans
+would create dangerous deletion pressure.
+
+## Design
+
+Serato Doctor now separates two read-only findings:
+
+- **Not in crates**: scanned tracks whose normalized filename is absent from all
+  loaded crate references. These may still be ordinary Serato library tracks.
+- **Possible orphan files**: tracks absent by normalized filename from both
+  loaded crates and database V2.
+
+When database V2 is unavailable, orphan status is reported as not assessed—not
+zero. Orphan candidates include preview and Finder controls but no delete or
+repair action.
+
+## Edge cases
+
+- Filename comparison is case- and Unicode-normalized.
+- A track in several folders may still require duplicate review separately.
+- Smart-crate materialized references count as crate evidence.
+- Missing database V2 prevents a defensible orphan conclusion.
+
+## Tests
+
+- Tracks known only to database V2 are excluded from orphan candidates.
+- Tracks in crates are excluded even if absent from database V2.
+- A track absent from both sources is counted once.
+- No database produces a not-assessed result.

serato_doctor/duplicates.py

@@ -11,7 +11,9 @@ def find_duplicate_groups(
 ) -> Tuple[DuplicateGroup, ...]:
     """Find exact-name duplicates and suspected numeric conflict copies."""
 
-    track_tuple = tuple(tracks)
+    # Healthy symlinks preserve legacy Serato paths without consuming another
+    # copy of the audio, so they are aliases rather than duplicate files.
+    track_tuple = tuple(track for track in tracks if not track.path.is_symlink())
     by_name: DefaultDict[str, List[DiskTrack]] = defaultdict(list)
     by_conflict_name: DefaultDict[str, List[DiskTrack]] = defaultdict(list)
 

serato_doctor/hashing.py

@@ -0,0 +1,33 @@
+"""Targeted content hashing for duplicate candidates."""
+
+import hashlib
+from pathlib import Path
+from typing import Iterable
+
+
+def sha256_file(path: Path, chunk_size: int = 1024 * 1024) -> str:
+    digest = hashlib.sha256()
+    with path.open("rb") as source:
+        for chunk in iter(lambda: source.read(chunk_size), b""):
+            digest.update(chunk)
+    return digest.hexdigest()
+
+
+def compare_audio_files(paths: Iterable[Path]) -> dict:
+    paths = tuple(paths)
+    if len(paths) < 2:
+        raise ValueError("At least two files are required for comparison")
+    fingerprints = []
+    for path in paths:
+        fingerprints.append(
+            {
+                "path": str(path),
+                "size": path.stat().st_size,
+                "sha256": sha256_file(path),
+            }
+        )
+    identical = len({item["sha256"] for item in fingerprints}) == 1
+    return {
+        "status": "identical" if identical else "different",
+        "files": fingerprints,
+    }

serato_doctor/health.py

@@ -36,9 +36,13 @@ def analyze_health(library: Library) -> HealthReport:
         for group in duplicate_groups
         if group.kind is DuplicateKind.CLOUD_CONFLICT
     ]
-    referenced_names = {reference.filename for reference in library.references}
+    referenced_names = {
+        normalize(reference.filename) for reference in library.references
+    }
     unused_count = sum(
-        1 for track in library.tracks if track.filename not in referenced_names
+        1
+        for track in library.tracks
+        if normalize(track.filename) not in referenced_names
     )
 
     matcher = MatchingEngine(library.tracks)
@@ -73,6 +77,15 @@ def analyze_health(library: Library) -> HealthReport:
             group.extra_files for group in cloud_conflicts
         ),
         unused_tracks=unused_count,
+        orphan_candidates=(
+            sum(
+                normalize(track.filename) not in referenced_names
+                and normalize(track.filename) not in database_names
+                for track in library.tracks
+            )
+            if library.database
+            else None
+        ),
         suggested_matches=suggested_count,
         broken_symlinks=len(library.broken_symlinks),
         static_crates=sum(

serato_doctor/models/health.py

@@ -18,6 +18,7 @@ class HealthReport:
     suspected_cloud_conflict_groups: int
     suspected_cloud_conflict_files: int
     unused_tracks: int
+    orphan_candidates: Optional[int]
     suggested_matches: int
     broken_symlinks: int
     static_crates: int

serato_doctor/repair.py

@@ -0,0 +1,242 @@
+"""Backup-first duplicate consolidation.
+
+Serato's binary metadata is deliberately not rewritten.  Removed duplicate files
+are replaced with symbolic links, so every existing path continues to resolve.
+"""
+
+import json
+import shutil
+from dataclasses import dataclass
+from datetime import datetime, timezone
+from pathlib import Path
+from typing import Iterable, Optional, Tuple
+from uuid import uuid4
+
+
+BACKUP_FOLDER = ".serato-doctor-backups"
+
+
+@dataclass(frozen=True)
+class DuplicateRepairPlan:
+    keeper: Path
+    replaced: Tuple[Path, ...]
+    metadata_files: Tuple[Path, ...]
+
+    @property
+    def changes(self) -> int:
+        return len(self.replaced)
+
+
+@dataclass(frozen=True)
+class RepairReceipt:
+    backup: Path
+    keeper: Path
+    replaced: Tuple[Path, ...]
+
+
+@dataclass(frozen=True)
+class BackupSummary:
+    path: Path
+    created_at: str
+    keeper: Path
+    replaced: Tuple[Path, ...]
+    size_bytes: int
+    restored_at: Optional[str]
+
+    @property
+    def status(self) -> str:
+        return "restored" if self.restored_at else "ready"
+
+
+def plan_duplicate_repair(
+    keeper: Path, duplicates: Iterable[Path], serato_root: Path
+) -> DuplicateRepairPlan:
+    keeper = keeper.expanduser().resolve()
+    candidates = tuple(path.expanduser().resolve() for path in duplicates)
+    if keeper not in candidates:
+        raise ValueError("The file to keep must belong to this duplicate group")
+    if not keeper.is_file():
+        raise ValueError(f"The file to keep no longer exists: {keeper}")
+    replaced = tuple(path for path in candidates if path != keeper)
+    if not replaced:
+        raise ValueError("Choose a duplicate group containing at least two files")
+    if any(not path.is_file() or path.is_symlink() for path in replaced):
+        raise ValueError("A duplicate changed since the analysis; analyze again")
+    metadata = tuple(
+        sorted(
+            path
+            for path in serato_root.expanduser().resolve().rglob("*")
+            if path.is_file()
+            and BACKUP_FOLDER not in path.parts
+            and (
+                path.name == "database V2"
+                or path.suffix.casefold() in {".crate", ".scrate"}
+            )
+        )
+    )
+    return DuplicateRepairPlan(keeper, replaced, metadata)
+
+
+def apply_duplicate_repair(
+    plan: DuplicateRepairPlan,
+    serato_root: Path,
+    backup_limit: Optional[int] = 10,
+) -> RepairReceipt:
+    """Create a complete rollback snapshot, then replace extras with symlinks."""
+    return apply_duplicate_repair_batch((plan,), serato_root, backup_limit)
+
+
+def apply_duplicate_repair_batch(
+    plans: Iterable[DuplicateRepairPlan],
+    serato_root: Path,
+    backup_limit: Optional[int] = 10,
+) -> RepairReceipt:
+    """Apply several approved duplicate choices as one atomic backup."""
+    plans = tuple(plans)
+    if not plans:
+        raise ValueError("Choose at least one duplicate group")
+    if backup_limit is not None and backup_limit < 1:
+        raise ValueError("Backup limit must be at least 1, or unlimited")
+    replaced_paths = tuple(
+        path for plan in plans for path in plan.replaced
+    )
+    if len(set(replaced_paths)) != len(replaced_paths):
+        raise ValueError("The same file appears in more than one repair choice")
+    keepers = {plan.keeper for plan in plans}
+    if keepers.intersection(replaced_paths):
+        raise ValueError("A selected keeper cannot be removed by another choice")
+    backup_root = serato_root.expanduser().resolve() / BACKUP_FOLDER
+    backup = backup_root / _backup_name()
+    files_root = backup / "files"
+    metadata_root = backup / "serato-metadata"
+    backup.mkdir(parents=True)
+
+    entries = []
+    try:
+        for plan in plans:
+            for path in plan.replaced:
+                destination = files_root / _safe_backup_path(path)
+                destination.parent.mkdir(parents=True, exist_ok=True)
+                shutil.copy2(path, destination)
+                entries.append(
+                    {
+                        "original": str(path),
+                        "backup": str(destination),
+                        "keeper": str(plan.keeper),
+                    }
+                )
+        for path in plans[0].metadata_files:
+            relative = path.relative_to(serato_root.expanduser().resolve())
+            destination = metadata_root / relative
+            destination.parent.mkdir(parents=True, exist_ok=True)
+            shutil.copy2(path, destination)
+        manifest = {
+            "created_at": datetime.now(timezone.utc).isoformat(),
+            "keeper": str(plans[0].keeper),
+            "keepers": [str(plan.keeper) for plan in plans],
+            "choice_count": len(plans),
+            "replaced": entries,
+            "strategy": "symlink",
+        }
+        (backup / "manifest.json").write_text(
+            json.dumps(manifest, indent=2), encoding="utf-8"
+        )
+        for entry in entries:
+            path = Path(entry["original"])
+            path.unlink()
+            path.symlink_to(Path(entry["keeper"]))
+    except Exception:
+        _rollback_entries(entries)
+        shutil.rmtree(backup, ignore_errors=True)
+        raise
+
+    rotate_backups(backup_root, backup_limit)
+    return RepairReceipt(backup, plans[0].keeper, replaced_paths)
+
+
+def restore_backup(backup: Path) -> Tuple[Path, ...]:
+    manifest_path = backup.expanduser().resolve() / "manifest.json"
+    manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
+    restored = []
+    for entry in manifest["replaced"]:
+        original = Path(entry["original"])
+        saved = Path(entry["backup"])
+        if original.exists() and not original.is_symlink():
+            raise ValueError(f"Restore would overwrite a real file: {original}")
+        if original.is_symlink():
+            original.unlink()
+        original.parent.mkdir(parents=True, exist_ok=True)
+        shutil.copy2(saved, original)
+        restored.append(original)
+    manifest["restored_at"] = datetime.now(timezone.utc).isoformat()
+    manifest_path.write_text(json.dumps(manifest, indent=2), encoding="utf-8")
+    return tuple(restored)
+
+
+def list_backups(serato_root: Path) -> Tuple[BackupSummary, ...]:
+    backup_root = serato_root.expanduser().resolve() / BACKUP_FOLDER
+    if not backup_root.is_dir():
+        return ()
+    summaries = []
+    for backup in backup_root.iterdir():
+        manifest_path = backup / "manifest.json"
+        if not manifest_path.is_file():
+            continue
+        try:
+            manifest = json.loads(manifest_path.read_text(encoding="utf-8"))
+            replaced = tuple(
+                Path(entry["original"]) for entry in manifest["replaced"]
+            )
+            size = sum(
+                path.stat().st_size
+                for path in backup.rglob("*")
+                if path.is_file()
+            )
+            summaries.append(
+                BackupSummary(
+                    path=backup.resolve(),
+                    created_at=manifest["created_at"],
+                    keeper=Path(manifest["keeper"]),
+                    replaced=replaced,
+                    size_bytes=size,
+                    restored_at=manifest.get("restored_at"),
+                )
+            )
+        except (KeyError, OSError, TypeError, json.JSONDecodeError):
+            continue
+    return tuple(
+        sorted(summaries, key=lambda item: item.created_at, reverse=True)
+    )
+
+
+def rotate_backups(backup_root: Path, limit: Optional[int]) -> None:
+    if limit is None or not backup_root.is_dir():
+        return
+    backups = sorted(
+        (path for path in backup_root.iterdir() if (path / "manifest.json").is_file()),
+        key=lambda path: path.name,
+        reverse=True,
+    )
+    for expired in backups[limit:]:
+        shutil.rmtree(expired)
+
+
+def _backup_name() -> str:
+    stamp = datetime.now(timezone.utc).strftime("%Y%m%dT%H%M%SZ")
+    return f"{stamp}-{uuid4().hex[:8]}"
+
+
+def _safe_backup_path(path: Path) -> Path:
+    anchorless = path.as_posix().lstrip("/").replace(":", "_")
+    return Path(anchorless)
+
+
+def _rollback_entries(entries: Iterable[dict]) -> None:
+    for entry in entries:
+        original = Path(entry["original"])
+        saved = Path(entry["backup"])
+        if original.is_symlink():
+            original.unlink()
+        if not original.exists() and saved.is_file():
+            original.parent.mkdir(parents=True, exist_ok=True)
+            shutil.copy2(saved, original)

serato_doctor/web.py

@@ -1,26 +1,297 @@
 import argparse
+import base64
+import binascii
+import hmac
 import json
+import mimetypes
+import secrets
+import subprocess
+from collections import Counter
 from dataclasses import asdict
 from http.server import BaseHTTPRequestHandler, ThreadingHTTPServer
 from importlib import resources
 from pathlib import Path
-from typing import Iterable
+from typing import Iterable, Optional
+from urllib.parse import parse_qs, quote, urlsplit
 
 from serato_doctor.crate_parser import load_library_crates
 from serato_doctor.database_parser import parse_database
+from serato_doctor.duplicates import find_duplicate_groups
 from serato_doctor.health import analyze_health
+from serato_doctor.hashing import compare_audio_files
+from serato_doctor.matching import MatchingEngine, normalize
+from serato_doctor.models.crate import CrateKind
+from serato_doctor.models.duplicate import DuplicateKind
 from serato_doctor.models.library import Library
 from serato_doctor.scanner import scan_filesystem
+from serato_doctor.repair import (
+    BACKUP_FOLDER,
+    apply_duplicate_repair,
+    apply_duplicate_repair_batch,
+    list_backups,
+    plan_duplicate_repair,
+    restore_backup,
+)
 
 
 MAX_REQUEST_BYTES = 64 * 1024
+DETAIL_LIMIT = 50
+FILE_TOKEN_SECRET = secrets.token_bytes(32)
 STATIC_FILES = {
     "/": ("index.html", "text/html; charset=utf-8"),
     "/app.css": ("app.css", "text/css; charset=utf-8"),
+    "/recovery.css": ("recovery.css", "text/css; charset=utf-8"),
+    "/layout-fixes.css": ("layout-fixes.css", "text/css; charset=utf-8"),
     "/app.js": ("app.js", "text/javascript; charset=utf-8"),
 }
 
 
+def _display_path(path: Path) -> str:
+    return str(path)
+
+
+def _file_token(path: Path) -> str:
+    encoded = base64.urlsafe_b64encode(str(path.resolve()).encode()).decode()
+    signature = hmac.digest(FILE_TOKEN_SECRET, encoded.encode(), "sha256").hex()
+    return f"{encoded}.{signature}"
+
+
+def _verified_audio(token: str) -> Path:
+    try:
+        encoded, signature = token.rsplit(".", 1)
+        expected = hmac.digest(
+            FILE_TOKEN_SECRET, encoded.encode(), "sha256"
+        ).hex()
+        if not hmac.compare_digest(signature, expected):
+            raise ValueError
+        path = Path(base64.urlsafe_b64decode(encoded.encode()).decode())
+    except (binascii.Error, ValueError, UnicodeDecodeError):
+        raise ValueError("Invalid or expired file preview")
+    if not path.is_file() or path.suffix.casefold() not in {
+        ".mp3",
+        ".m4a",
+        ".wav",
+        ".aif",
+        ".aiff",
+        ".flac",
+    }:
+        raise ValueError("Audio file is no longer available")
+    return path
+
+
+def _file_preview(path: Path) -> dict:
+    token = _file_token(path)
+    return {
+        "path": _display_path(path),
+        "audio_url": f"/api/audio?token={quote(token)}",
+        "reveal_token": token,
+    }
+
+
+def _first_reason(match) -> str:
+    for item in match.evidence:
+        if item.matched:
+            return item.explanation
+    return "Filename-related candidate"
+
+
+def diagnostic_details(library: Library, limit: int = DETAIL_LIMIT) -> dict:
+    dynamic_sources = {
+        crate.path for crate in library.crates if crate.kind is CrateKind.SMART
+    }
+    results = tuple(
+        result
+        for result in library.reconcile_by_filename()
+        if result.reference.source not in dynamic_sources
+    )
+    missing = [result for result in results if not result.exists_by_filename]
+
+    matcher = MatchingEngine(library.tracks)
+    suggested_matches = []
+    for result in missing:
+        candidates = matcher.candidates_for(result.reference)
+        if not candidates:
+            continue
+        best = candidates[0]
+        suggested_matches.append(
+            {
+                "filename": result.reference.filename,
+                "crate": _display_path(result.reference.source),
+                "saved_path": _display_path(result.reference.path),
+                "candidate": _display_path(best.track.path),
+                "score": f"{best.score_percent}%",
+                "reason": _first_reason(best),
+            }
+        )
+
+    duplicate_groups = find_duplicate_groups(library.tracks)
+    exact_duplicates = [
+        group
+        for group in duplicate_groups
+        if group.kind is DuplicateKind.EXACT_NAME
+    ]
+    cloud_conflicts = [
+        group
+        for group in duplicate_groups
+        if group.kind is DuplicateKind.CLOUD_CONFLICT
+    ]
+
+    database_tracks = library.database.tracks if library.database else ()
+    missing_database_tracks = [
+        track for track in database_tracks if not track.path.exists()
+    ]
+    database_filename_counts = Counter(
+        normalize(track.filename) for track in missing_database_tracks
+    )
+    referenced_names = {
+        normalize(reference.filename) for reference in library.references
+    }
+    database_names = {normalize(track.filename) for track in database_tracks}
+    unused_tracks = [
+        track
+        for track in library.tracks
+        if normalize(track.filename) not in referenced_names
+    ]
+    orphan_candidates = (
+        [
+            track
+            for track in unused_tracks
+            if normalize(track.filename) not in database_names
+        ]
+        if library.database
+        else None
+    )
+
+    return {
+        "database_missing_tracks": {
+            "title": "Missing tracks in Serato",
+            "summary": (
+                "These are Serato database entries whose saved file location "
+                "does not currently exist on disk."
+            ),
+            "total": len(missing_database_tracks),
+            "items": [
+                {
+                    "filename": track.filename,
+                    "saved_path": _display_path(track.path),
+                    "artist": track.artist or "Unknown artist",
+                    "title": track.title or track.filename,
+                    "repeated_filename": database_filename_counts[
+                        normalize(track.filename)
+                    ]
+                    > 1,
+                }
+                for track in missing_database_tracks[:limit]
+            ],
+        },
+        "old_crate_references": {
+            "title": "Old crate references",
+            "summary": (
+                "These are regular crate appearances whose exact filename was "
+                "not found in the selected music folder."
+            ),
+            "total": len(missing),
+            "items": [
+                {
+                    "filename": result.reference.filename,
+                    "crate": _display_path(result.reference.source),
+                    "saved_path": _display_path(result.reference.path),
+                }
+                for result in missing[:limit]
+            ],
+        },
+        "suggested_matches": {
+            "title": "Suggested matches",
+            "summary": (
+                "These are read-only guesses where Serato Doctor found a "
+                "filename-related candidate on disk."
+            ),
+            "total": len(suggested_matches),
+            "items": suggested_matches[:limit],
+        },
+        "duplicate_filenames": {
+            "title": "Duplicate filenames",
+            "summary": (
+                "These groups contain different files with the same cleaned-up "
+                "filename. Review before making any decisions."
+            ),
+            "total": len(exact_duplicates),
+            "items": [
+                {
+                    "filename": group.display_name,
+                    "files": [_display_path(track.path) for track in group.tracks],
+                    "file_previews": [
+                        _file_preview(track.path) for track in group.tracks
+                    ],
+                }
+                for group in exact_duplicates[:limit]
+            ],
+        },
+        "cloud_conflicts": {
+            "title": "Possible cloud conflicts",
+            "summary": (
+                "These filename families look like cloud sync conflict copies, "
+                "such as a duplicate ending in a number."
+            ),
+            "total": len(cloud_conflicts),
+            "items": [
+                {
+                    "filename": group.display_name,
+                    "files": [_display_path(track.path) for track in group.tracks],
+                    "file_previews": [
+                        _file_preview(track.path) for track in group.tracks
+                    ],
+                }
+                for group in cloud_conflicts[:limit]
+            ],
+        },
+        "broken_symlinks": {
+            "title": "Broken shortcuts",
+            "summary": (
+                "These symbolic links point somewhere that no longer resolves."
+            ),
+            "total": len(library.broken_symlinks),
+            "items": [
+                {
+                    "path": _display_path(link.path),
+                    "target": _display_path(link.target) if link.target else "Unknown",
+                }
+                for link in library.broken_symlinks[:limit]
+            ],
+        },
+        "unused_tracks": {
+            "title": "Not in crates",
+            "summary": (
+                "These files are not in any loaded crate, but may still be "
+                "normal tracks in Serato's main library."
+            ),
+            "total": len(unused_tracks),
+            "items": [
+                {"filename": track.filename, "path": _display_path(track.path)}
+                for track in unused_tracks[:limit]
+            ],
+        },
+        "orphan_candidates": {
+            "title": "Possible orphan files",
+            "summary": (
+                "These files were found in the music folder but not by filename "
+                "in loaded crates or Serato's database. Review only; this is "
+                "never an automatic deletion recommendation."
+            ),
+            "total": len(orphan_candidates) if orphan_candidates is not None else None,
+            "assessed": orphan_candidates is not None,
+            "items": [
+                {
+                    "filename": track.filename,
+                    "path": _display_path(track.path),
+                    **_file_preview(track.path),
+                }
+                for track in (orphan_candidates or ())[:limit]
+            ],
+        },
+    }
+
+
 def analyze_paths(
     serato: Path, music: Path, reference_roots: Iterable[Path] = ()
 ) -> dict:
@@ -45,12 +316,142 @@ def analyze_paths(
     report = analyze_health(library)
     result = asdict(report)
     result["score_basis"] = report.score_basis
+    result["details"] = diagnostic_details(library)
     return result
 
 
+def duplicate_repair(
+    serato: Path,
+    music: Path,
+    keeper: Path,
+    group_files: Iterable[Path],
+    backup_limit: Optional[int],
+    apply: bool = False,
+) -> dict:
+    """Validate a current duplicate group and preview or apply consolidation."""
+    serato = serato.expanduser().resolve()
+    music = music.expanduser().resolve()
+    keeper = keeper.expanduser().resolve()
+    requested = tuple(path.expanduser().resolve() for path in group_files)
+    if not serato.is_dir() or not music.is_dir():
+        raise ValueError("Analyze the library again before repairing duplicates")
+    tracks = scan_filesystem(music).tracks
+    groups = find_duplicate_groups(tracks)
+    valid_groups = [
+        {track.path.resolve() for track in group.tracks} for group in groups
+    ]
+    if set(requested) not in valid_groups:
+        raise ValueError("This duplicate group changed; analyze the library again")
+    plan = plan_duplicate_repair(keeper, requested, serato)
+    result = {
+        "keeper": str(plan.keeper),
+        "replaced": [str(path) for path in plan.replaced],
+        "metadata_backups": len(plan.metadata_files),
+        "strategy": "shortcut",
+        "database_v2_modified": False,
+    }
+    if apply:
+        receipt = apply_duplicate_repair(plan, serato, backup_limit)
+        result.update({"applied": True, "backup": str(receipt.backup)})
+    else:
+        result["applied"] = False
+    return result
+
+
+def duplicate_repair_batch(
+    serato: Path,
+    music: Path,
+    choices: Iterable[dict],
+    backup_limit: Optional[int],
+    apply: bool = False,
+) -> dict:
+    """Validate and preview or apply several keeper choices together."""
+    serato = serato.expanduser().resolve()
+    music = music.expanduser().resolve()
+    if not serato.is_dir() or not music.is_dir():
+        raise ValueError("Analyze the library again before repairing duplicates")
+    groups = find_duplicate_groups(scan_filesystem(music).tracks)
+    valid_groups = [
+        {track.path.resolve() for track in group.tracks} for group in groups
+    ]
+    plans = []
+    selected_groups = set()
+    for choice in choices:
+        requested = tuple(Path(value).expanduser().resolve() for value in choice["group_files"])
+        group_key = frozenset(requested)
+        if set(requested) not in valid_groups:
+            raise ValueError("A duplicate group changed; analyze the library again")
+        if group_key in selected_groups:
+            raise ValueError("A duplicate group was selected more than once")
+        selected_groups.add(group_key)
+        plans.append(
+            plan_duplicate_repair(Path(choice["keeper"]), requested, serato)
+        )
+    if not plans:
+        raise ValueError("Choose at least one duplicate group")
+    replaced = [str(path) for plan in plans for path in plan.replaced]
+    comparisons = [
+        compare_audio_files((plan.keeper,) + plan.replaced) for plan in plans
+    ]
+    result = {
+        "choice_count": len(plans),
+        "replaced": replaced,
+        "decisions": [
+            {
+                "keeper": str(plan.keeper),
+                "replaced": [str(path) for path in plan.replaced],
+                "hash_status": comparison["status"],
+            }
+            for plan, comparison in zip(plans, comparisons)
+        ],
+        "metadata_backups": len(plans[0].metadata_files),
+        "strategy": "shortcut",
+        "database_v2_modified": False,
+    }
+    if apply:
+        receipt = apply_duplicate_repair_batch(plans, serato, backup_limit)
+        result.update({"applied": True, "backup": str(receipt.backup)})
+    else:
+        result["applied"] = False
+    return result
+
+
+def backup_history(serato: Path) -> dict:
+    serato = serato.expanduser().resolve()
+    if not serato.is_dir():
+        raise ValueError(f"Serato folder does not exist: {serato}")
+    backups = list_backups(serato)
+    return {
+        "total": len(backups),
+        "size_bytes": sum(backup.size_bytes for backup in backups),
+        "items": [
+            {
+                "path": str(backup.path),
+                "created_at": backup.created_at,
+                "keeper": str(backup.keeper),
+                "replaced": [str(path) for path in backup.replaced],
+                "size_bytes": backup.size_bytes,
+                "status": backup.status,
+                "restored_at": backup.restored_at,
+            }
+            for backup in backups
+        ],
+    }
+
+
 class SeratoDoctorHandler(BaseHTTPRequestHandler):
     def do_GET(self) -> None:
-        asset = STATIC_FILES.get(self.path)
+        request = urlsplit(self.path)
+        if request.path == "/api/audio":
+            try:
+                token = parse_qs(request.query)["token"][0]
+                self._audio_response(_verified_audio(token))
+            except (BrokenPipeError, ConnectionResetError):
+                return
+            except (KeyError, IndexError, OSError, ValueError) as error:
+                self._json_response(404, {"error": str(error)})
+            return
+        asset = STATIC_FILES.get(request.path)
         if asset is None:
             self._json_response(404, {"error": "Not found"})
             return
@@ -62,12 +463,24 @@ class SeratoDoctorHandler(BaseHTTPRequestHandler):
         )
         self.send_response(200)
         self.send_header("Content-Type", content_type)
+        self.send_header("Cache-Control", "no-store")
         self.send_header("Content-Length", str(len(content)))
         self.end_headers()
         self.wfile.write(content)
 
     def do_POST(self) -> None:
-        if self.path != "/api/analyze":
+        allowed = {
+            "/api/analyze",
+            "/api/duplicates/preview",
+            "/api/duplicates/apply",
+            "/api/duplicates/batch/preview",
+            "/api/duplicates/batch/apply",
+            "/api/duplicates/hash",
+            "/api/backups/restore",
+            "/api/backups",
+            "/api/reveal",
+        }
+        if self.path not in allowed:
             self._json_response(404, {"error": "Not found"})
             return
         try:
@@ -77,15 +490,107 @@ class SeratoDoctorHandler(BaseHTTPRequestHandler):
             payload = json.loads(self.rfile.read(length))
             if not isinstance(payload, dict):
                 raise ValueError("Request body must be a JSON object")
-            roots = [Path(value) for value in payload.get("reference_roots", [])]
+            if self.path == "/api/analyze":
-            result = analyze_paths(
+                roots = [Path(value) for value in payload.get("reference_roots", [])]
-                Path(payload["serato"]), Path(payload["music"]), roots
+                result = analyze_paths(
-            )
+                    Path(payload["serato"]), Path(payload["music"]), roots
-        except (KeyError, TypeError, json.JSONDecodeError, ValueError) as error:
+                )
+            elif self.path == "/api/reveal":
+                path = _verified_audio(payload["token"])
+                subprocess.run(["open", "-R", str(path)], check=True)
+                result = {"revealed": str(path)}
+            elif self.path == "/api/backups":
+                result = backup_history(Path(payload["serato"]))
+            elif self.path == "/api/backups/restore":
+                serato = Path(payload["serato"]).expanduser().resolve()
+                backup = Path(payload["backup"]).expanduser().resolve()
+                backup_root = (serato / BACKUP_FOLDER).resolve()
+                if backup.parent != backup_root:
+                    raise ValueError("That backup does not belong to this library")
+                restored = restore_backup(backup)
+                result = {"restored": [str(path) for path in restored]}
+            elif self.path == "/api/duplicates/hash":
+                tokens = payload["tokens"]
+                if not isinstance(tokens, list) or not 2 <= len(tokens) <= 10:
+                    raise ValueError("Compare between 2 and 10 audio files")
+                comparison = compare_audio_files(
+                    _verified_audio(token) for token in tokens
+                )
+                result = {
+                    "status": comparison["status"],
+                    "files": [
+                        {
+                            "path": item["path"],
+                            "size": item["size"],
+                            "fingerprint": item["sha256"][:12],
+                        }
+                        for item in comparison["files"]
+                    ],
+                }
+            elif self.path.startswith("/api/duplicates/batch/"):
+                raw_limit = payload.get("backup_limit", 10)
+                backup_limit = None if raw_limit is None else int(raw_limit)
+                result = duplicate_repair_batch(
+                    Path(payload["serato"]),
+                    Path(payload["music"]),
+                    payload["choices"],
+                    backup_limit,
+                    apply=self.path.endswith("/apply"),
+                )
+            else:
+                raw_limit = payload.get("backup_limit", 10)
+                backup_limit = None if raw_limit is None else int(raw_limit)
+                result = duplicate_repair(
+                    Path(payload["serato"]),
+                    Path(payload["music"]),
+                    Path(payload["keeper"]),
+                    (Path(value) for value in payload["group_files"]),
+                    backup_limit,
+                    apply=self.path.endswith("/apply"),
+                )
+        except (
+            KeyError,
+            TypeError,
+            OSError,
+            json.JSONDecodeError,
+            ValueError,
+        ) as error:
             self._json_response(400, {"error": str(error)})
             return
         self._json_response(200, result)
 
+    def _audio_response(self, path: Path) -> None:
+        size = path.stat().st_size
+        start, end = 0, size - 1
+        status = 200
+        range_header = self.headers.get("Range")
+        if range_header and range_header.startswith("bytes="):
+            raw_start, _, raw_end = range_header[6:].partition("-")
+            start = int(raw_start or 0)
+            end = min(int(raw_end) if raw_end else end, end)
+            if start < 0 or start > end:
+                raise ValueError("Invalid audio range")
+            status = 206
+        length = end - start + 1
+        self.send_response(status)
+        self.send_header(
+            "Content-Type", mimetypes.guess_type(path.name)[0] or "audio/mpeg"
+        )
+        self.send_header("Accept-Ranges", "bytes")
+        self.send_header("Content-Length", str(length))
+        if status == 206:
+            self.send_header("Content-Range", f"bytes {start}-{end}/{size}")
+        self.end_headers()
+        with path.open("rb") as audio:
+            audio.seek(start)
+            remaining = length
+            while remaining:
+                chunk = audio.read(min(64 * 1024, remaining))
+                if not chunk:
+                    break
+                self.wfile.write(chunk)
+                remaining -= len(chunk)
+
     def _json_response(self, status: int, payload: dict) -> None:
         content = json.dumps(payload).encode("utf-8")
         self.send_response(status)

serato_doctor/webui/app.js

@@ -3,12 +3,330 @@ const button = document.querySelector('#analyze-button');
 const errorBox = document.querySelector('#error-message');
 const results = document.querySelector('#dashboard');
 const infoButtons = document.querySelectorAll('.info-button');
+const drillTriggers = document.querySelectorAll('[data-detail]');
+const drilldownTitle = document.querySelector('#drilldown-title');
+const drilldownCount = document.querySelector('#drilldown-count');
+const drilldownSummary = document.querySelector('#drilldown-summary');
+const drilldownList = document.querySelector('#drilldown-list');
+const repairPanel = document.querySelector('#duplicate-repair');
+const repairChoice = document.querySelector('#repair-choice');
+const repairPreview = document.querySelector('#repair-preview');
+const repairMessage = document.querySelector('#repair-message');
+const previewRepairButton = document.querySelector('#preview-repair');
+const applyRepairButton = document.querySelector('#apply-repair');
+const restoreRepairButton = document.querySelector('#restore-repair');
+const loadBackupsButton = document.querySelector('#load-backups');
+const backupSummary = document.querySelector('#backup-summary');
+const backupList = document.querySelector('#backup-list');
+const audioPreview = document.querySelector('#audio-preview');
+const audioPreviewName = document.querySelector('#audio-preview-name');
+const audioPlayer = document.querySelector('#audio-player');
+const closeAudioPreview = document.querySelector('#close-audio-preview');
+const batchPreviewModal = document.querySelector('#batch-preview-modal');
+const batchPreviewSummary = document.querySelector('#batch-preview-summary');
+const batchPreviewList = document.querySelector('#batch-preview-list');
+const batchPreviewSafety = document.querySelector('#batch-preview-safety');
+const closeBatchPreview = document.querySelector('#close-batch-preview');
+const continueReviewing = document.querySelector('#continue-reviewing');
+const acceptPreview = document.querySelector('#accept-preview');
+let latestAnalysis = null;
+let selectedDuplicateGroup = null;
+let previewedRepair = null;
+let latestBackup = null;
+let reviewState = null;
 
 function expandHome(path) {
   return path.trim();
 }
 
+function escapeHtml(value) {
+  return String(value ?? '').replace(/[&<>"']/g, (character) => ({
+    '&': '&amp;',
+    '<': '&lt;',
+    '>': '&gt;',
+    '"': '&quot;',
+    "'": '&#039;',
+  }[character]));
+}
+
+function formatBytes(bytes) {
+  if (!bytes) return '0 B';
+  const units = ['B', 'KB', 'MB', 'GB', 'TB'];
+  const unit = Math.min(Math.floor(Math.log(bytes) / Math.log(1024)), units.length - 1);
+  return `${(bytes / (1024 ** unit)).toFixed(unit ? 1 : 0)} ${units[unit]}`;
+}
+
+function formatDate(value) {
+  const date = new Date(value);
+  return Number.isNaN(date.valueOf()) ? value : date.toLocaleString([], {dateStyle: 'medium', timeStyle: 'short'});
+}
+
+async function loadBackups() {
+  loadBackupsButton.disabled = true;
+  backupSummary.textContent = 'Looking for recovery snapshots…';
+  try {
+    const response = await fetch('/api/backups', {method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({serato: expandHome(document.querySelector('#serato-path').value)})});
+    const data = await response.json();
+    if (!response.ok) throw new Error(data.error || 'Could not load backups');
+    backupSummary.textContent = `${data.total} backup${data.total === 1 ? '' : 's'} · ${formatBytes(data.size_bytes)} on disk`;
+    backupList.innerHTML = data.items.length ? data.items.map((backup) => `
+      <article class="backup-item">
+        <div><span class="backup-status ${backup.status}">${backup.status === 'restored' ? 'Restored' : 'Ready to restore'}</span><strong>${formatDate(backup.created_at)}</strong><small>Kept: ${escapeHtml(backup.keeper)}</small><small>${backup.replaced.length} original file${backup.replaced.length === 1 ? '' : 's'} · ${formatBytes(backup.size_bytes)}</small></div>
+        <button type="button" data-restore-backup="${escapeHtml(backup.path)}" ${backup.status === 'restored' ? 'disabled' : ''}>${backup.status === 'restored' ? 'Already restored' : 'Restore'}</button>
+      </article>
+    `).join('') : '<div class="empty-detail">No repair backups found for this library yet.</div>';
+  } catch (error) {
+    backupSummary.textContent = error.message;
+    backupList.innerHTML = '';
+  } finally { loadBackupsButton.disabled = false; }
+}
+
+loadBackupsButton.addEventListener('click', loadBackups);
+backupList.addEventListener('click', async (event) => {
+  const button = event.target.closest('[data-restore-backup]');
+  if (!button || button.disabled) return;
+  if (!window.confirm('Restore the original duplicate files from this backup? Existing real files will never be overwritten.')) return;
+  button.disabled = true; button.textContent = 'Restoring…';
+  try {
+    const response = await fetch('/api/backups/restore', {method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({serato: expandHome(document.querySelector('#serato-path').value), backup: button.dataset.restoreBackup})});
+    const result = await response.json();
+    if (!response.ok) throw new Error(result.error || 'Restore failed');
+    await loadBackups();
+  } catch (error) { backupSummary.textContent = error.message; button.disabled = false; button.textContent = 'Restore'; }
+});
+
+function detailLines(item) {
+  if (item.file_previews) {
+    return item.file_previews.map((file) => `<li class="file-compare-row"><span>${escapeHtml(file.path)}</span><div><button type="button" data-audio-url="${escapeHtml(file.audio_url)}" data-audio-name="${escapeHtml(file.path)}">▶ Play preview</button><button type="button" data-reveal-token="${escapeHtml(file.reveal_token)}">Show in Finder</button></div></li>`).join('');
+  }
+  const lines = [];
+  if (item.artist || item.title) lines.push(`${item.artist || 'Unknown artist'} — ${item.title || item.filename}`);
+  if (item.crate) lines.push(`Crate: ${item.crate}`);
+  if (item.saved_path) lines.push(`Saved path: ${item.saved_path}`);
+  if (item.candidate) lines.push(`Candidate: ${item.candidate}`);
+  if (item.path) lines.push(`File: ${item.path}`);
+  if (item.target) lines.push(`Target: ${item.target}`);
+  if (item.score || item.reason) lines.push(`${item.score || 'Match'} · ${item.reason || 'Candidate found'}`);
+  if (item.repeated_filename) lines.push('Same filename appears more than once in Serato’s missing list.');
+  return lines.map((line) => `<li>${escapeHtml(line)}</li>`).join('');
+}
+
+function detailActions(item) {
+  if (!item.audio_url || !item.reveal_token) return '';
+  return `<div class="file-actions detail-actions"><button type="button" data-audio-url="${escapeHtml(item.audio_url)}" data-audio-name="${escapeHtml(item.path)}">▶ Play preview</button><button type="button" data-reveal-token="${escapeHtml(item.reveal_token)}">Show in Finder</button></div>`;
+}
+
+function renderDetail(key) {
+  const detail = latestAnalysis?.details?.[key];
+  if (!detail) return;
+
+  drillTriggers.forEach((trigger) => {
+    trigger.classList.toggle('selected', trigger.dataset.detail === key);
+  });
+  drilldownTitle.textContent = detail.title;
+  drilldownCount.textContent = detail.assessed === false ? 'Not assessed' : `${detail.total ?? 0} found`;
+  drilldownSummary.textContent = detail.summary;
+
+  repairPanel.hidden = true;
+  selectedDuplicateGroup = null;
+  previewedRepair = null;
+  latestBackup = null;
+  restoreRepairButton.hidden = true;
+  if (detail.assessed === false) {
+    drilldownList.innerHTML = '<div class="empty-detail">Serato’s database V2 was not available, so orphan status cannot be assessed safely.</div>';
+    return;
+  }
+  if (!detail.items?.length) {
+    drilldownList.innerHTML = '<div class="empty-detail">Nothing to review here. Tiny victory parade, very tasteful.</div>';
+    return;
+  }
+
+  if (detail.items[0].file_previews) {
+    reviewState = {key, index: 0, choices: new Map()};
+    repairPanel.hidden = false;
+    renderReviewGroup();
+    return;
+  }
+
+  drilldownList.innerHTML = detail.items.map((item, index) => `
+    <article class="detail-item">
+      <strong>${escapeHtml(item.filename || item.path || 'Untitled item')}</strong>
+      <ul>${detailLines(item)}</ul>
+      ${detailActions(item)}
+    </article>
+  `).join('');
+}
+
+function renderReviewGroup() {
+  const groups = latestAnalysis.details[reviewState.key].items;
+  const group = groups[reviewState.index];
+  const chosen = reviewState.choices.get(reviewState.index);
+  drilldownCount.textContent = `${reviewState.index + 1} of ${groups.length} · ${reviewState.choices.size} approved`;
+  drilldownSummary.textContent = 'Listen to each candidate, choose the keeper, and we’ll move to the next group. Skip anything uncertain.';
+  drilldownList.innerHTML = `
+    <article class="review-workspace">
+      <div class="review-heading"><div><span>Comparing now</span><strong>${escapeHtml(group.filename)}</strong></div><div class="review-signals"><span id="hash-confidence" class="hash-confidence checking">Checking file identity…</span><span>${reviewState.choices.size} selected</span></div></div>
+      <div class="review-candidates">${group.file_previews.map((file, index) => `
+        <section class="review-candidate ${chosen === file.path ? 'winner' : ''}">
+          <span class="candidate-number">Option ${index + 1}</span>
+          <strong>${escapeHtml(file.path.split('/').pop())}</strong>
+          <small>${escapeHtml(file.path)}</small>
+          <div class="file-actions"><button type="button" data-audio-url="${escapeHtml(file.audio_url)}" data-audio-name="${escapeHtml(file.path)}">▶ Play preview</button><button type="button" data-reveal-token="${escapeHtml(file.reveal_token)}">Show in Finder</button></div>
+          <button class="choose-winner" type="button" data-choose-winner="${index}">${chosen === file.path ? '✓ Selected keeper' : 'Keep this one →'}</button>
+        </section>
+      `).join('')}</div>
+      <div class="review-navigation"><button type="button" data-review-previous ${reviewState.index === 0 ? 'disabled' : ''}>← Previous</button><button type="button" data-review-skip>Skip for now</button></div>
+    </article>`;
+  updateBatchSummary();
+  loadHashConfidence(group, reviewState.index, reviewState.key);
+}
+
+async function loadHashConfidence(group, groupIndex, detailKey) {
+  const badge = document.querySelector('#hash-confidence');
+  if (group.hashComparison) {
+    showHashConfidence(badge, group.hashComparison);
+    return;
+  }
+  try {
+    const response = await fetch('/api/duplicates/hash', {method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({tokens: group.file_previews.map((file) => file.reveal_token)})});
+    const result = await response.json();
+    if (!response.ok) throw new Error(result.error || 'Identity check failed');
+    group.hashComparison = result;
+    if (reviewState?.index === groupIndex && reviewState?.key === detailKey) showHashConfidence(document.querySelector('#hash-confidence'), result);
+  } catch (error) {
+    if (reviewState?.index === groupIndex && reviewState?.key === detailKey && badge) { badge.className = 'hash-confidence unavailable'; badge.textContent = 'Identity not verified'; }
+  }
+}
+
+function showHashConfidence(badge, comparison) {
+  if (!badge) return;
+  badge.className = `hash-confidence ${comparison.status}`;
+  badge.textContent = comparison.status === 'identical' ? '✓ Byte-for-byte identical' : '⚠ Files differ — listen carefully';
+  badge.title = comparison.files.map((file) => `${formatBytes(file.size)} · ${file.fingerprint}`).join('\n');
+}
+
+function updateBatchSummary() {
+  if (!reviewState) return;
+  const groups = latestAnalysis.details[reviewState.key].items;
+  repairChoice.innerHTML = `<div class="batch-summary"><strong>${reviewState.choices.size} group${reviewState.choices.size === 1 ? '' : 's'} approved</strong><span>${groups.length - reviewState.choices.size} skipped or still awaiting a decision</span></div>`;
+  previewRepairButton.disabled = reviewState.choices.size === 0;
+  previewedRepair = null;
+  applyRepairButton.disabled = true;
+  repairPreview.hidden = true;
+}
+
+function repairPayload() {
+  if (!reviewState?.choices.size) throw new Error('Choose at least one keeper');
+  const groups = latestAnalysis.details[reviewState.key].items;
+  const choices = Array.from(reviewState.choices, ([index, keeper]) => ({keeper, group_files: groups[index].files}));
+  return {serato: expandHome(document.querySelector('#serato-path').value), music: expandHome(document.querySelector('#music-path').value), choices, backup_limit: document.querySelector('#keep-all-backups').checked ? null : Number(document.querySelector('#backup-limit').value)};
+}
+
+async function requestRepair(endpoint) {
+  const response = await fetch(endpoint, {method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify(repairPayload())});
+  const data = await response.json();
+  if (!response.ok) throw new Error(data.error || 'Duplicate cleanup failed');
+  return data;
+}
+
+previewRepairButton.addEventListener('click', async () => {
+  repairMessage.textContent = 'Checking the plan…'; applyRepairButton.disabled = true;
+  try {
+    previewedRepair = await requestRepair('/api/duplicates/batch/preview');
+    batchPreviewSummary.textContent = `${previewedRepair.choice_count} keeper decision(s) · ${previewedRepair.replaced.length} duplicate file(s) consolidated`;
+    batchPreviewList.innerHTML = previewedRepair.decisions.map((decision, index) => `
+      <article class="preview-decision"><span>Decision ${index + 1}<em class="hash-confidence ${decision.hash_status}">${decision.hash_status === 'identical' ? 'Identical' : 'Files differ'}</em></span><div class="winner-path"><b>Keep</b><strong>${escapeHtml(decision.keeper.split('/').pop())}</strong><small>${escapeHtml(decision.keeper)}</small></div><div class="replaced-paths"><b>Replace with a shortcut</b>${decision.replaced.map((path) => `<small>${escapeHtml(path)}</small>`).join('')}</div></article>
+    `).join('');
+    batchPreviewSafety.textContent = `${previewedRepair.metadata_backups} Serato metadata file(s) and every replaced audio file will be backed up. Database V2 will not be modified.`;
+    repairPreview.innerHTML = `<strong>Preview approved</strong><p>${previewedRepair.choice_count} keeper decision(s) are ready for one backed-up apply.</p>`;
+    repairPreview.hidden = false; applyRepairButton.disabled = false;
+    repairMessage.textContent = 'Preview complete. Nothing has changed yet.';
+    batchPreviewModal.showModal();
+  } catch (error) { repairMessage.textContent = error.message; }
+});
+
+function dismissBatchPreview() { batchPreviewModal.close(); }
+closeBatchPreview.addEventListener('click', dismissBatchPreview);
+continueReviewing.addEventListener('click', dismissBatchPreview);
+acceptPreview.addEventListener('click', dismissBatchPreview);
+
+applyRepairButton.addEventListener('click', async () => {
+  if (!previewedRepair) return;
+  applyRepairButton.disabled = true; repairMessage.textContent = 'Creating the backup before making changes…';
+  try {
+    const result = await requestRepair('/api/duplicates/batch/apply');
+    latestBackup = result.backup;
+    repairMessage.textContent = `Cleanup complete. Restore backup: ${result.backup}`;
+    previewRepairButton.disabled = true;
+    restoreRepairButton.hidden = false;
+    await loadBackups();
+  } catch (error) { repairMessage.textContent = error.message; applyRepairButton.disabled = false; }
+});
+
+restoreRepairButton.addEventListener('click', async () => {
+  if (!latestBackup) return;
+  restoreRepairButton.disabled = true; repairMessage.textContent = 'Restoring the duplicate files…';
+  try {
+    const response = await fetch('/api/backups/restore', {method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({serato: expandHome(document.querySelector('#serato-path').value), backup: latestBackup})});
+    const result = await response.json();
+    if (!response.ok) throw new Error(result.error || 'Restore failed');
+    repairMessage.textContent = `Restore complete. ${result.restored.length} original file(s) returned.`;
+    restoreRepairButton.hidden = true;
+    await loadBackups();
+  } catch (error) { repairMessage.textContent = error.message; restoreRepairButton.disabled = false; }
+});
+
+async function handleFileAction(event) {
+  const previewButton = event.target.closest('[data-audio-url]');
+  const revealButton = event.target.closest('[data-reveal-token]');
+  if (!previewButton && !revealButton) return false;
+  event.preventDefault(); event.stopPropagation();
+  if (previewButton) {
+    audioPlayer.src = previewButton.dataset.audioUrl;
+    audioPreviewName.textContent = previewButton.dataset.audioName.split('/').pop();
+    audioPreview.hidden = false;
+    try { await audioPlayer.play(); } catch (_) { /* Native controls remain available. */ }
+  } else {
+    const original = revealButton.textContent;
+    revealButton.disabled = true; revealButton.textContent = 'Opening…';
+    try {
+      const response = await fetch('/api/reveal', {method: 'POST', headers: {'Content-Type': 'application/json'}, body: JSON.stringify({token: revealButton.dataset.revealToken})});
+      const result = await response.json();
+      if (!response.ok) throw new Error(result.error || 'Could not open Finder');
+      revealButton.textContent = 'Shown in Finder';
+    } catch (error) { revealButton.textContent = error.message; }
+    setTimeout(() => { revealButton.disabled = false; revealButton.textContent = original; }, 1800);
+  }
+  return true;
+}
+
+closeAudioPreview.addEventListener('click', () => { audioPlayer.pause(); audioPlayer.removeAttribute('src'); audioPlayer.load(); audioPreview.hidden = true; });
+repairChoice.addEventListener('click', handleFileAction);
+drilldownList.addEventListener('click', async (event) => {
+  if (await handleFileAction(event) || !reviewState) return;
+  const groups = latestAnalysis.details[reviewState.key].items;
+  const winner = event.target.closest('[data-choose-winner]');
+  if (winner) {
+    const fileIndex = Number(winner.dataset.chooseWinner);
+    reviewState.choices.set(reviewState.index, groups[reviewState.index].files[fileIndex]);
+    if (reviewState.index < groups.length - 1) reviewState.index += 1;
+    renderReviewGroup();
+    return;
+  }
+  if (event.target.closest('[data-review-skip]')) {
+    reviewState.choices.delete(reviewState.index);
+    if (reviewState.index < groups.length - 1) reviewState.index += 1;
+    renderReviewGroup();
+    return;
+  }
+  if (event.target.closest('[data-review-previous]') && reviewState.index > 0) {
+    reviewState.index -= 1;
+    renderReviewGroup();
+  }
+});
+
 function render(data) {
+  latestAnalysis = data;
   document.querySelectorAll('[data-field]').forEach((element) => {
     const value = data[element.dataset.field];
     element.textContent = value ?? '—';
@@ -21,6 +339,7 @@ function render(data) {
     : score >= 95 ? 'Looking excellent' : score >= 80 ? 'A few things need attention' : 'Review recommended';
   document.querySelector('#score-basis').textContent = data.score_basis;
   document.querySelector('#analysis-time').textContent = `Completed ${new Date().toLocaleTimeString([], {hour: '2-digit', minute: '2-digit'})}`;
+  renderDetail(data.database_missing_paths > 0 ? 'database_missing_tracks' : 'old_crate_references');
   results.hidden = false;
   results.scrollIntoView({behavior: 'smooth', block: 'start'});
 }
@@ -44,6 +363,16 @@ infoButtons.forEach((infoButton) => {
   });
 });
 
+drillTriggers.forEach((trigger) => {
+  trigger.addEventListener('click', () => renderDetail(trigger.dataset.detail));
+  trigger.addEventListener('keydown', (event) => {
+    if (event.key === 'Enter' || event.key === ' ') {
+      event.preventDefault();
+      renderDetail(trigger.dataset.detail);
+    }
+  });
+});
+
 document.addEventListener('click', () => closeInfoButtons());
 document.addEventListener('keydown', (event) => {
   if (event.key === 'Escape') closeInfoButtons();

serato_doctor/webui/index.html

@@ -5,7 +5,9 @@
   <meta name="viewport" content="width=device-width, initial-scale=1">
   <meta name="color-scheme" content="dark">
   <title>Serato Doctor</title>
-  <link rel="stylesheet" href="/app.css">
+  <link rel="stylesheet" href="/app.css?v=3">
+  <link rel="stylesheet" href="/recovery.css?v=3">
+  <link rel="stylesheet" href="/layout-fixes.css?v=3">
 </head>
 <body>
   <div class="ambient ambient-one"></div>
@@ -20,10 +22,11 @@
         <a class="nav-item active" href="#dashboard"><span>⌁</span> Dashboard</a>
         <a class="nav-item" href="#scan"><span>◎</span> New analysis</a>
         <a class="nav-item" href="#diagnostics"><span>◇</span> Diagnostics</a>
+        <a class="nav-item" href="#recovery"><span>↶</span> Recovery</a>
       </nav>
       <div class="safety-card">
         <span class="safety-icon">✓</span>
-        <div><strong>Read-only mode</strong><p>Your library will not be modified.</p></div>
+        <div><strong>Protected mode</strong><p>Analysis is read-only. Repairs require a preview and backup.</p></div>
       </div>
       <div class="sidebar-foot">Local interface · v0.1</div>
     </aside>
@@ -65,9 +68,9 @@
           </article>
           <div class="metrics-grid">
             <article class="metric panel"><button class="info-button" type="button" aria-label="About tracks" aria-expanded="false" data-info="Audio files found inside the music folder you selected. This is the collection Serato Doctor compared with your crates and database.">i</button><span>Tracks scanned</span><strong data-field="disk_tracks">—</strong><small>audio files found</small></article>
-            <article class="metric panel warning"><button class="info-button" type="button" aria-label="About missing tracks in Serato" aria-expanded="false" data-info="Tracks in Serato's database whose saved file location no longer exists. This should be close to the orange or unmapped track count you see in Serato.">i</button><span>Missing tracks in Serato</span><strong data-field="database_missing_paths">—</strong><small><b data-field="database_missing_unique_filenames">—</b> unique filenames</small></article>
+            <article class="metric panel warning drill-trigger" role="button" tabindex="0" data-detail="database_missing_tracks"><button class="info-button" type="button" aria-label="About missing tracks in Serato" aria-expanded="false" data-info="Tracks in Serato's database whose saved file location no longer exists. This should be close to the orange or unmapped track count you see in Serato.">i</button><span>Missing tracks in Serato</span><strong data-field="database_missing_paths">—</strong><small><b data-field="database_missing_unique_filenames">—</b> unique filenames · click for list</small></article>
-            <article class="metric panel"><button class="info-button" type="button" aria-label="About unused tracks" aria-expanded="false" data-info="Files in the selected music folder whose filename is not used by any loaded crate. They may still be valid library tracks; this is informational, not a deletion recommendation.">i</button><span>Unused tracks</span><strong data-field="unused_tracks">—</strong><small>not referenced by crates</small></article>
+            <article class="metric panel drill-trigger" role="button" tabindex="0" data-detail="orphan_candidates"><button class="info-button" type="button" aria-label="About possible orphan files" aria-expanded="false" data-info="Files found in the music folder but not by filename in loaded crates or Serato's database. They require review and are never automatic deletion recommendations.">i</button><span>Possible orphan files</span><strong data-field="orphan_candidates">—</strong><small>outside crates and database · click for list</small></article>
-            <article class="metric panel"><button class="info-button" type="button" aria-label="About suggested matches" aria-expanded="false" data-info="Missing crate entries with a filename-related candidate, such as an added OneDrive conflict number. Suggestions are evidence for review, never automatic repairs.">i</button><span>Suggested matches</span><strong data-field="suggested_matches">—</strong><small>explainable candidates</small></article>
+            <article class="metric panel drill-trigger" role="button" tabindex="0" data-detail="suggested_matches"><button class="info-button" type="button" aria-label="About suggested matches" aria-expanded="false" data-info="Missing crate entries with a filename-related candidate, such as an added OneDrive conflict number. Suggestions are evidence for review, never automatic repairs.">i</button><span>Suggested matches</span><strong data-field="suggested_matches">—</strong><small>explainable candidates · click for list</small></article>
           </div>
         </div>
 
@@ -75,16 +78,59 @@
           <div class="section-heading"><div><p class="eyebrow">Full picture</p><h2>Diagnostics</h2></div><span class="read-only-tag">No changes made</span></div>
           <div class="diagnostic-list">
             <div><span class="diag-icon violet">◇</span><p><strong>Crates</strong><small><b data-field="static_crates">—</b> regular · <b data-field="smart_crates">—</b> smart · <b data-field="smart_crate_containers">—</b> dynamic containers</small></p><button class="info-button" type="button" aria-label="About crates" aria-expanded="false" data-info="Regular crates are lists you maintain by hand. Smart crates are rebuilt by Serato from rules, so their generated references are not scored as broken.">i</button></div>
-            <div><span class="diag-icon amber">≋</span><p><strong>Duplicate filenames</strong><small><b data-field="duplicate_filename_groups">—</b> exact groups · <b data-field="duplicate_files">—</b> extra files</small></p><button class="info-button" type="button" aria-label="About duplicate filenames" aria-expanded="false" data-info="Different files with the same filename after case and Unicode cleanup. They need review, but matching names alone do not mean either file should be deleted.">i</button></div>
+            <div class="drill-trigger" role="button" tabindex="0" data-detail="duplicate_filenames"><span class="diag-icon amber">≋</span><p><strong>Duplicate filenames</strong><small><b data-field="duplicate_filename_groups">—</b> exact groups · <b data-field="duplicate_files">—</b> extra files · click for list</small></p><button class="info-button" type="button" aria-label="About duplicate filenames" aria-expanded="false" data-info="Different files with the same filename after case and Unicode cleanup. They need review, but matching names alone do not mean either file should be deleted.">i</button></div>
-            <div><span class="diag-icon blue">⌁</span><p><strong>Possible cloud conflicts</strong><small><b data-field="suspected_cloud_conflict_groups">—</b> groups · <b data-field="suspected_cloud_conflict_files">—</b> extra files</small></p><button class="info-button" type="button" aria-label="About cloud conflicts" aria-expanded="false" data-info="Filename families such as Track.mp3 and Track 2.mp3. OneDrive often creates these during sync conflicts, but numbered song titles can also be legitimate.">i</button></div>
+            <div class="drill-trigger" role="button" tabindex="0" data-detail="cloud_conflicts"><span class="diag-icon blue">⌁</span><p><strong>Possible cloud conflicts</strong><small><b data-field="suspected_cloud_conflict_groups">—</b> groups · <b data-field="suspected_cloud_conflict_files">—</b> extra files · click for list</small></p><button class="info-button" type="button" aria-label="About cloud conflicts" aria-expanded="false" data-info="Filename families such as Track.mp3 and Track 2.mp3. OneDrive often creates these during sync conflicts, but numbered song titles can also be legitimate.">i</button></div>
-            <div><span class="diag-icon red">↗</span><p><strong>Broken shortcuts</strong><small><b data-field="broken_symlinks">—</b> unresolved symbolic links</small></p><button class="info-button" type="button" aria-label="About broken shortcuts" aria-expanded="false" data-info="Shortcut-style symbolic links whose destination no longer exists. Serato Doctor reports them but never removes or recreates them automatically.">i</button></div>
+            <div class="drill-trigger" role="button" tabindex="0" data-detail="broken_symlinks"><span class="diag-icon red">↗</span><p><strong>Broken shortcuts</strong><small><b data-field="broken_symlinks">—</b> unresolved symbolic links · click for list</small></p><button class="info-button" type="button" aria-label="About broken shortcuts" aria-expanded="false" data-info="Shortcut-style symbolic links whose destination no longer exists. Serato Doctor reports them but never removes or recreates them automatically.">i</button></div>
-            <div><span class="diag-icon violet">▦</span><p><strong>Old crate references</strong><small><b data-field="missing_references">—</b> appearances · <b data-field="unique_missing_filenames">—</b> unique filenames</small></p><button class="info-button" type="button" aria-label="About old crate references" aria-expanded="false" data-info="Saved spots in regular crates whose exact filename was not found in the selected music folder. The same track can appear in several crates, so appearances are higher than unique filenames. These are separate from Serato's unmapped-track count.">i</button></div>
+            <div class="drill-trigger" role="button" tabindex="0" data-detail="old_crate_references"><span class="diag-icon violet">▦</span><p><strong>Old crate references</strong><small><b data-field="missing_references">—</b> appearances · <b data-field="unique_missing_filenames">—</b> unique filenames · click for list</small></p><button class="info-button" type="button" aria-label="About old crate references" aria-expanded="false" data-info="Saved spots in regular crates whose exact filename was not found in the selected music folder. The same track can appear in several crates, so appearances are higher than unique filenames. These are separate from Serato's unmapped-track count.">i</button></div>
+            <div class="drill-trigger" role="button" tabindex="0" data-detail="unused_tracks"><span class="diag-icon blue">◌</span><p><strong>Not in crates</strong><small><b data-field="unused_tracks">—</b> tracks still potentially in Serato · click for list</small></p><button class="info-button" type="button" aria-label="About tracks not in crates" aria-expanded="false" data-info="Tracks that do not appear in loaded crates. They may still be normal, active tracks in Serato's main library and should not be treated as orphans.">i</button></div>
             <div><span class="diag-icon amber">⌕</span><p><strong>Database coverage</strong><small><b data-field="database_entries">—</b> Serato entries · <b data-field="tracks_missing_from_database">—</b> scanned tracks absent</small></p><button class="info-button" type="button" aria-label="About database coverage" aria-expanded="false" data-info="Compares filenames in Serato's database with the selected music folder. A scanned track absent from the database may not have been imported, or may be represented under another filename.">i</button></div>
           </div>
         </div>
+
+        <div id="drilldowns" class="drilldowns panel">
+          <div class="section-heading"><div><p class="eyebrow">Look closer</p><h2 id="drilldown-title">Choose a diagnostic</h2></div><span id="drilldown-count">Read-only examples</span></div>
+          <p id="drilldown-summary">Click a metric above to see example files and saved paths behind that number.</p>
+          <div id="drilldown-list" class="detail-list"></div>
+        </div>
+
+        <div id="duplicate-repair" class="repair-panel panel" hidden>
+          <div class="section-heading"><div><p class="eyebrow">Batch cleanup</p><h2>Review all selected changes</h2></div><span class="repair-tag">One backup · one apply</span></div>
+          <p>Approved choices are combined into one plan. Skipped groups remain untouched, and Serato’s database V2 is never edited.</p>
+          <div id="repair-choice" class="repair-choice"></div>
+          <div class="backup-options">
+            <label>Backups to keep <input id="backup-limit" type="number" min="1" value="10"></label>
+            <label class="check-label"><input id="keep-all-backups" type="checkbox"> Keep every backup</label>
+          </div>
+          <div id="repair-preview" class="repair-preview" hidden></div>
+          <div class="repair-actions">
+            <button id="preview-repair" type="button">Preview all selected changes</button>
+            <button id="apply-repair" class="danger-action" type="button" disabled>Apply all selected changes</button>
+            <button id="restore-repair" type="button" hidden>Restore this backup</button>
+          </div>
+          <div id="repair-message" class="repair-message" role="status"></div>
+        </div>
       </section>
+
+      <section id="recovery" class="recovery-panel panel">
+        <div class="section-heading"><div><p class="eyebrow">Safety net</p><h2>Backup recovery</h2></div><button id="load-backups" type="button">Check backups</button></div>
+        <p>Repair backups appear here automatically before any duplicate is consolidated. Review saved snapshots, disk usage, and restore history.</p>
+        <div id="backup-summary" class="backup-summary">Analyze your library, then choose a duplicate to preview its automatic backup.</div>
+        <div id="backup-list" class="backup-list"></div>
+      </section>
+      <div id="audio-preview" class="audio-preview" hidden>
+        <div><span>Now previewing</span><strong id="audio-preview-name"></strong></div>
+        <audio id="audio-player" controls preload="metadata"></audio>
+        <button id="close-audio-preview" type="button" aria-label="Close audio preview">×</button>
+      </div>
+      <dialog id="batch-preview-modal" class="batch-preview-modal">
+        <div class="modal-heading"><div><p class="eyebrow">Final dry-run</p><h2>Here’s exactly what will happen</h2></div><button id="close-batch-preview" type="button" aria-label="Close preview">×</button></div>
+        <p id="batch-preview-summary"></p>
+        <div id="batch-preview-list" class="batch-preview-list"></div>
+        <div class="modal-safety"><span>✓</span><p><strong>Backup happens first</strong><small id="batch-preview-safety"></small></p></div>
+        <div class="modal-actions"><button id="continue-reviewing" type="button">Close and continue reviewing</button><button id="accept-preview" type="button">Looks right</button></div>
+      </dialog>
     </main>
   </div>
-  <script src="/app.js" defer></script>
+  <script src="/app.js?v=7" defer></script>
 </body>
 </html>

serato_doctor/webui/layout-fixes.css

@@ -0,0 +1,447 @@
+.recovery-panel {
+  margin-top: 42px;
+}
+
+@media (max-width: 680px) {
+  .recovery-panel {
+    margin-top: 28px;
+  }
+}
+
+.review-workspace {
+  padding: clamp(16px, 3vw, 24px);
+  border: 1px solid rgba(155, 135, 245, .18);
+  border-radius: 16px;
+  background: rgba(255, 255, 255, .02);
+}
+
+.review-heading,
+.review-navigation {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 16px;
+}
+
+.review-heading span,
+.batch-summary span {
+  color: var(--muted);
+  font-size: 10px;
+}
+
+.review-heading strong {
+  display: block;
+  margin-top: 5px;
+  font-size: 14px;
+}
+
+.review-signals {
+  display: flex;
+  align-items: flex-end;
+  gap: 7px;
+  flex-direction: column;
+}
+
+.hash-confidence {
+  display: inline-block;
+  padding: 5px 8px;
+  border-radius: 999px;
+  font-size: 9px!important;
+  font-style: normal;
+  font-weight: 750;
+  letter-spacing: .02em!important;
+  text-transform: none!important;
+}
+
+.hash-confidence.checking,
+.hash-confidence.unavailable {
+  color: var(--muted);
+  background: rgba(255, 255, 255, .05);
+}
+
+.hash-confidence.identical {
+  color: #8ce4b8;
+  background: rgba(84, 212, 154, .11);
+}
+
+.hash-confidence.different {
+  color: var(--amber);
+  background: rgba(245, 185, 76, .1);
+}
+
+.preview-decision > span .hash-confidence {
+  display: block;
+  width: max-content;
+  margin-top: 8px;
+}
+
+.review-candidates {
+  display: grid;
+  grid-template-columns: repeat(2, minmax(0, 1fr));
+  gap: 12px;
+  margin: 18px 0;
+}
+
+.review-candidate {
+  display: flex;
+  flex-direction: column;
+  gap: 9px;
+  min-width: 0;
+  padding: 16px;
+  border: 1px solid var(--line);
+  border-radius: 14px;
+  background: rgba(6, 8, 12, .28);
+}
+
+.review-candidate.winner {
+  border-color: rgba(84, 212, 154, .52);
+  background: rgba(84, 212, 154, .06);
+}
+
+.candidate-number {
+  color: var(--violet);
+  font-size: 9px;
+  font-weight: 750;
+  text-transform: uppercase;
+  letter-spacing: .1em;
+}
+
+.review-candidate > strong,
+.review-candidate > small {
+  overflow-wrap: anywhere;
+}
+
+.review-candidate > strong {
+  font-size: 12px;
+}
+
+.review-candidate > small {
+  flex: 1;
+  color: var(--muted);
+  font-size: 9px;
+  line-height: 1.5;
+}
+
+.choose-winner {
+  margin-top: 4px;
+  border: 0;
+  border-radius: 10px;
+  padding: 10px 12px;
+  background: linear-gradient(135deg, #907be9, #6e59cf);
+  color: white;
+  font: 750 11px/1 inherit;
+  cursor: pointer;
+}
+
+.winner .choose-winner {
+  background: rgba(84, 212, 154, .18);
+  color: #8ce4b8;
+}
+
+.review-navigation button {
+  border: 1px solid var(--line);
+  border-radius: 9px;
+  padding: 8px 11px;
+  background: rgba(255, 255, 255, .04);
+  color: var(--muted);
+  font: 700 10px/1 inherit;
+  cursor: pointer;
+}
+
+.review-navigation button:disabled {
+  opacity: .35;
+}
+
+.batch-summary {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 12px;
+  padding: 14px;
+  border: 1px solid rgba(102, 217, 232, .2);
+  border-radius: 12px;
+  background: rgba(102, 217, 232, .05);
+}
+
+.batch-summary strong {
+  font-size: 12px;
+}
+
+@media (max-width: 760px) {
+  .review-candidates {
+    grid-template-columns: 1fr;
+  }
+
+  .review-heading,
+  .batch-summary {
+    align-items: flex-start;
+    flex-direction: column;
+  }
+}
+
+.batch-preview-modal {
+  width: min(820px, calc(100vw - 32px));
+  max-height: min(82vh, 760px);
+  padding: 24px;
+  overflow: auto;
+  border: 1px solid rgba(155, 135, 245, .3);
+  border-radius: 20px;
+  background: #161923;
+  color: var(--text);
+  box-shadow: 0 30px 100px rgba(0, 0, 0, .65);
+}
+
+.batch-preview-modal::backdrop {
+  background: rgba(4, 5, 8, .78);
+  backdrop-filter: blur(7px);
+}
+
+.modal-heading,
+.modal-actions,
+.modal-safety {
+  display: flex;
+  align-items: center;
+  justify-content: space-between;
+  gap: 14px;
+}
+
+.modal-heading h2 {
+  margin: 0;
+  font-size: 23px;
+  letter-spacing: -.03em;
+}
+
+.modal-heading > button {
+  border: 0;
+  background: transparent;
+  color: var(--muted);
+  font-size: 28px;
+  cursor: pointer;
+}
+
+#batch-preview-summary {
+  color: var(--muted);
+  font-size: 12px;
+}
+
+.batch-preview-list {
+  display: grid;
+  gap: 10px;
+  max-height: 390px;
+  margin: 18px 0;
+  padding-right: 4px;
+  overflow: auto;
+}
+
+.preview-decision {
+  display: grid;
+  grid-template-columns: 80px 1fr 1fr;
+  gap: 13px;
+  padding: 14px;
+  border: 1px solid var(--line);
+  border-radius: 13px;
+  background: rgba(255, 255, 255, .025);
+}
+
+.preview-decision > span,
+.preview-decision b {
+  color: var(--muted);
+  font-size: 9px;
+  text-transform: uppercase;
+  letter-spacing: .08em;
+}
+
+.winner-path,
+.replaced-paths {
+  min-width: 0;
+}
+
+.winner-path strong,
+.winner-path small,
+.replaced-paths small {
+  display: block;
+  margin-top: 5px;
+  overflow-wrap: anywhere;
+}
+
+.winner-path strong {
+  color: #8ce4b8;
+  font-size: 11px;
+}
+
+.winner-path small,
+.replaced-paths small {
+  color: var(--muted);
+  font-size: 9px;
+  line-height: 1.45;
+}
+
+.modal-safety {
+  justify-content: flex-start;
+  padding: 13px;
+  border: 1px solid rgba(102, 217, 232, .2);
+  border-radius: 12px;
+  background: rgba(102, 217, 232, .05);
+}
+
+.modal-safety > span {
+  color: var(--cyan);
+}
+
+.modal-safety p,
+.modal-safety strong,
+.modal-safety small {
+  display: block;
+  margin: 0;
+}
+
+.modal-safety strong {
+  font-size: 11px;
+}
+
+.modal-safety small {
+  margin-top: 4px;
+  color: var(--muted);
+  font-size: 9px;
+  line-height: 1.45;
+}
+
+.modal-actions {
+  justify-content: flex-end;
+  margin-top: 16px;
+}
+
+.modal-actions button {
+  border: 1px solid var(--line);
+  border-radius: 10px;
+  padding: 10px 13px;
+  background: rgba(255, 255, 255, .04);
+  color: var(--text);
+  font: 700 11px/1 inherit;
+  cursor: pointer;
+}
+
+#accept-preview {
+  border-color: transparent;
+  background: #6f5bd0;
+}
+
+@media (max-width: 620px) {
+  .preview-decision {
+    grid-template-columns: 1fr;
+  }
+
+  .modal-actions {
+    align-items: stretch;
+    flex-direction: column-reverse;
+  }
+}
+
+.file-compare-row {
+  display: grid;
+  gap: 8px;
+  padding: 9px 0;
+  border-top: 1px solid rgba(255, 255, 255, .05);
+}
+
+.file-compare-row:first-child {
+  border-top: 0;
+}
+
+.file-compare-row div,
+.file-actions {
+  display: flex;
+  flex-wrap: wrap;
+  gap: 7px;
+}
+
+.file-compare-row button,
+.file-actions button {
+  border: 1px solid rgba(155, 135, 245, .25);
+  border-radius: 8px;
+  padding: 7px 9px;
+  background: rgba(155, 135, 245, .08);
+  color: #d8d2f6;
+  font: 700 10px/1 inherit;
+  cursor: pointer;
+}
+
+.keeper-option {
+  justify-content: space-between;
+}
+
+.keeper-option > label {
+  display: flex;
+  align-items: center;
+  gap: 12px;
+  min-width: 0;
+  cursor: pointer;
+}
+
+.audio-preview {
+  position: fixed;
+  right: 24px;
+  bottom: 20px;
+  display: grid;
+  grid-template-columns: minmax(150px, .7fr) minmax(240px, 1.3fr) auto;
+  align-items: center;
+  gap: 16px;
+  width: min(720px, calc(100vw - 48px));
+  padding: 13px 14px;
+  border: 1px solid rgba(102, 217, 232, .3);
+  border-radius: 14px;
+  background: rgba(16, 21, 29, .96);
+  box-shadow: 0 16px 48px rgba(0, 0, 0, .45);
+  z-index: 5;
+}
+
+.audio-preview[hidden] {
+  display: none;
+}
+
+.audio-preview span,
+.audio-preview strong {
+  display: block;
+}
+
+.audio-preview span {
+  color: var(--cyan);
+  font-size: 9px;
+  text-transform: uppercase;
+  letter-spacing: .1em;
+}
+
+.audio-preview strong {
+  max-width: 360px;
+  margin-top: 4px;
+  overflow: hidden;
+  color: var(--text);
+  font-size: 11px;
+  text-overflow: ellipsis;
+  white-space: nowrap;
+}
+
+.audio-preview audio {
+  width: 100%;
+  height: 34px;
+}
+
+.audio-preview > button {
+  border: 0;
+  background: transparent;
+  color: var(--muted);
+  font-size: 22px;
+  cursor: pointer;
+}
+
+@media (max-width: 680px) {
+  .keeper-option,
+  .audio-preview {
+    align-items: stretch;
+    grid-template-columns: 1fr;
+  }
+
+  .keeper-option {
+    flex-direction: column;
+  }
+}

serato_doctor/webui/recovery.css

@@ -0,0 +1 @@
+.recovery-panel{margin-top:18px;padding:25px}.recovery-panel>p{color:var(--muted);font-size:12px;line-height:1.6}.recovery-panel .section-heading button{border:1px solid rgba(155,135,245,.35);border-radius:10px;padding:9px 12px;background:rgba(155,135,245,.12);color:#ddd8fa;font:700 11px inherit;cursor:pointer}.recovery-panel button:disabled{opacity:.5;cursor:not-allowed}.backup-summary{padding:11px 13px;border-radius:11px;background:rgba(255,255,255,.03);color:#bfc4cf;font-size:11px;margin:15px 0 10px}.backup-list{display:grid;gap:9px}.backup-item{display:flex;align-items:center;justify-content:space-between;gap:18px;padding:14px;border:1px solid var(--line);border-radius:13px;background:rgba(255,255,255,.025)}.backup-item>div{min-width:0}.backup-item strong,.backup-item small{display:block}.backup-item strong{font-size:12px;margin:7px 0}.backup-item small{font-size:10px;color:var(--muted);line-height:1.5;overflow-wrap:anywhere}.backup-item>button{flex:0 0 auto;border:0;border-radius:10px;padding:10px 13px;background:#6f5bd0;color:#fff;font:700 11px inherit;cursor:pointer}.backup-status{display:inline-block;padding:4px 7px;border-radius:999px;font-size:9px;text-transform:uppercase;letter-spacing:.08em;color:var(--cyan);background:rgba(102,217,232,.1)}.backup-status.restored{color:#aeb4c2;background:rgba(255,255,255,.06)}@media(max-width:520px){.backup-item{align-items:stretch;flex-direction:column}.backup-item>button{width:100%}}

tests/test_hashing.py

@@ -0,0 +1,25 @@
+from serato_doctor.hashing import compare_audio_files, sha256_file
+
+
+def test_equal_files_have_the_same_hash(tmp_path):
+    first = tmp_path / "First.mp3"
+    second = tmp_path / "Second.mp3"
+    first.write_bytes(b"same audio bytes")
+    second.write_bytes(b"same audio bytes")
+
+    result = compare_audio_files((first, second))
+
+    assert result["status"] == "identical"
+    assert sha256_file(first) == sha256_file(second)
+
+
+def test_different_files_are_not_reported_as_identical(tmp_path):
+    first = tmp_path / "First.mp3"
+    second = tmp_path / "Second.mp3"
+    first.write_bytes(b"version one")
+    second.write_bytes(b"version two")
+
+    result = compare_audio_files((first, second))
+
+    assert result["status"] == "different"
+    assert result["files"][0]["sha256"] != result["files"][1]["sha256"]

tests/test_health.py

@@ -2,6 +2,7 @@ from pathlib import Path
 
 from serato_doctor.health import analyze_health
 from serato_doctor.models.crate import Crate, CrateKind
+from serato_doctor.models.database import DatabaseTrack, SeratoDatabase
 from serato_doctor.models.filesystem import BrokenSymlink
 from serato_doctor.models.library import Library
 from serato_doctor.models.reference import TrackReference
@@ -49,6 +50,7 @@ def test_health_report_exposes_each_metric():
     assert report.duplicate_filename_groups == 1
     assert report.duplicate_files == 1
     assert report.unused_tracks == 3
+    assert report.orphan_candidates is None
     assert report.suggested_matches == 1
 
 
@@ -60,6 +62,32 @@ def test_empty_library_has_no_health_score():
     assert report.scored_references == 0
     assert not report.database_present
     assert report.tracks_missing_from_database == 0
+    assert report.orphan_candidates is None
+
+
+def test_orphan_candidates_exclude_tracks_known_to_serato_or_crates():
+    tracks = [
+        track("Crated.mp3"),
+        track("LibraryOnly.mp3"),
+        track("OutsideEverything.mp3"),
+    ]
+    database = SeratoDatabase(
+        Path("database V2"),
+        "test",
+        (
+            DatabaseTrack(
+                Path("/new/House/LibraryOnly.mp3"), "LibraryOnly.mp3"
+            ),
+        ),
+    )
+    library = Library.build(
+        [reference("Crated.mp3")], tracks, database=database
+    )
+
+    report = analyze_health(library)
+
+    assert report.unused_tracks == 2
+    assert report.orphan_candidates == 1
 
 
 def test_smart_crate_references_are_reported_but_not_scored():

tests/test_repair.py

@@ -0,0 +1,101 @@
+from pathlib import Path
+
+import pytest
+
+from serato_doctor.repair import (
+    BACKUP_FOLDER,
+    apply_duplicate_repair,
+    apply_duplicate_repair_batch,
+    list_backups,
+    plan_duplicate_repair,
+    restore_backup,
+    rotate_backups,
+)
+
+
+def library(tmp_path):
+    serato = tmp_path / "_Serato_"
+    crate = serato / "Subcrates" / "House.crate"
+    crate.parent.mkdir(parents=True)
+    crate.write_bytes(b"crate-data")
+    (serato / "database V2").write_bytes(b"database-data")
+    music = tmp_path / "Music"
+    keeper = music / "Main" / "Track.mp3"
+    duplicate = music / "Old" / "Track.mp3"
+    keeper.parent.mkdir(parents=True)
+    duplicate.parent.mkdir(parents=True)
+    keeper.write_bytes(b"keeper")
+    duplicate.write_bytes(b"duplicate")
+    return serato, keeper, duplicate
+
+
+def test_duplicate_repair_backs_up_then_preserves_old_path_as_link(tmp_path):
+    serato, keeper, duplicate = library(tmp_path)
+    plan = plan_duplicate_repair(keeper, (keeper, duplicate), serato)
+
+    receipt = apply_duplicate_repair(plan, serato, backup_limit=3)
+
+    assert duplicate.is_symlink()
+    assert duplicate.resolve() == keeper.resolve()
+    assert (receipt.backup / "manifest.json").is_file()
+    assert any((receipt.backup / "serato-metadata").rglob("House.crate"))
+    assert any((receipt.backup / "serato-metadata").rglob("database V2"))
+
+    history = list_backups(serato)
+    assert len(history) == 1
+    assert history[0].status == "ready"
+    assert history[0].size_bytes > 0
+
+    restored = restore_backup(receipt.backup)
+    assert restored == (duplicate.resolve(),)
+    assert not duplicate.is_symlink()
+    assert duplicate.read_bytes() == b"duplicate"
+    assert list_backups(serato)[0].status == "restored"
+
+
+def test_plan_rejects_keeper_outside_duplicate_group(tmp_path):
+    serato, keeper, duplicate = library(tmp_path)
+    outsider = tmp_path / "outsider.mp3"
+    outsider.write_bytes(b"other")
+
+    with pytest.raises(ValueError, match="must belong"):
+        plan_duplicate_repair(outsider, (keeper, duplicate), serato)
+
+
+def test_backup_rotation_can_be_limited_or_unlimited(tmp_path):
+    root = tmp_path / BACKUP_FOLDER
+    for name in ("001", "002", "003"):
+        backup = root / name
+        backup.mkdir(parents=True)
+        (backup / "manifest.json").write_text("{}", encoding="utf-8")
+
+    rotate_backups(root, None)
+    assert len(list(root.iterdir())) == 3
+    rotate_backups(root, 2)
+    assert {path.name for path in root.iterdir()} == {"002", "003"}
+
+
+def test_batch_repair_uses_one_backup_and_restores_every_group(tmp_path):
+    serato, first_keeper, first_duplicate = library(tmp_path)
+    second_keeper = tmp_path / "Music" / "Main" / "Other.mp3"
+    second_duplicate = tmp_path / "Music" / "Old" / "Other.mp3"
+    second_keeper.write_bytes(b"other-keeper")
+    second_duplicate.write_bytes(b"other-duplicate")
+    plans = (
+        plan_duplicate_repair(
+            first_keeper, (first_keeper, first_duplicate), serato
+        ),
+        plan_duplicate_repair(
+            second_keeper, (second_keeper, second_duplicate), serato
+        ),
+    )
+
+    receipt = apply_duplicate_repair_batch(plans, serato, backup_limit=10)
+
+    assert first_duplicate.is_symlink()
+    assert second_duplicate.is_symlink()
+    assert len(list((serato / BACKUP_FOLDER).iterdir())) == 1
+    assert len(receipt.replaced) == 2
+    restore_backup(receipt.backup)
+    assert first_duplicate.read_bytes() == b"duplicate"
+    assert second_duplicate.read_bytes() == b"other-duplicate"

tests/test_web.py

@@ -3,7 +3,14 @@ from pathlib import Path
 
 import pytest
 
-from serato_doctor.web import STATIC_FILES, analyze_paths
+from serato_doctor.web import (
+    STATIC_FILES,
+    _file_token,
+    _verified_audio,
+    analyze_paths,
+    duplicate_repair,
+    duplicate_repair_batch,
+)
 
 
 def test_web_analysis_uses_production_health_pipeline(tmp_path):
@@ -26,6 +33,12 @@ def test_web_analysis_uses_production_health_pipeline(tmp_path):
     assert result["database_missing_paths"] == 0
     assert result["database_missing_unique_filenames"] == 0
     assert result["tracks_missing_from_database"] == 0
+    assert result["details"]["old_crate_references"]["total"] == 2
+    assert result["details"]["old_crate_references"]["items"][0]["crate"]
+    assert result["details"]["suggested_matches"]["total"] == 0
+    assert result["details"]["unused_tracks"]["total"] == 1
+    assert result["orphan_candidates"] == 0
+    assert result["details"]["orphan_candidates"]["assessed"] is True
 
 
 def test_web_analysis_rejects_missing_folders(tmp_path):
@@ -36,9 +49,103 @@ def test_web_analysis_rejects_missing_folders(tmp_path):
 def test_web_static_assets_are_declared_and_packaged():
     asset_root = Path(__file__).parents[1] / "serato_doctor" / "webui"
 
-    assert set(STATIC_FILES) == {"/", "/app.css", "/app.js"}
+    assert set(STATIC_FILES) == {
+        "/",
+        "/app.css",
+        "/recovery.css",
+        "/layout-fixes.css",
+        "/app.js",
+    }
     assert all((asset_root / filename).is_file() for filename, _ in STATIC_FILES.values())
     html = (asset_root / "index.html").read_text(encoding="utf-8")
     assert "Missing tracks in Serato" in html
     assert "Old crate references" in html
+    assert "Possible orphan files" in html
+    assert "Choose a diagnostic" in html
+    assert "Backup recovery" in html
+    assert "Here’s exactly what will happen" in html
+    assert 'id="batch-preview-modal"' in html
+    assert 'data-detail="database_missing_tracks"' in html
+    assert 'data-detail="old_crate_references"' in html
     assert html.count('class="info-button"') >= 10
+
+
+def test_duplicate_repair_preview_does_not_change_files(tmp_path):
+    serato = tmp_path / "_Serato_"
+    serato.mkdir()
+    music = tmp_path / "Music"
+    first = music / "A" / "Track.mp3"
+    second = music / "B" / "Track.mp3"
+    first.parent.mkdir(parents=True)
+    second.parent.mkdir(parents=True)
+    first.write_bytes(b"first")
+    second.write_bytes(b"second")
+
+    result = duplicate_repair(
+        serato, music, first, (first, second), backup_limit=10
+    )
+
+    assert result["applied"] is False
+    assert result["database_v2_modified"] is False
+    assert second.read_bytes() == b"second"
+    assert not second.is_symlink()
+
+
+def test_audio_preview_tokens_only_open_signed_audio(tmp_path):
+    track = tmp_path / "Track.mp3"
+    track.write_bytes(b"audio")
+
+    token = _file_token(track)
+
+    assert _verified_audio(token) == track
+    with pytest.raises(ValueError, match="Invalid or expired"):
+        _verified_audio(token + "changed")
+
+
+def test_duplicate_details_include_preview_and_finder_controls(tmp_path):
+    serato = tmp_path / "_Serato_"
+    serato.mkdir()
+    music = tmp_path / "Music"
+    first = music / "A" / "Track.mp3"
+    second = music / "B" / "Track.mp3"
+    first.parent.mkdir(parents=True)
+    second.parent.mkdir(parents=True)
+    first.write_bytes(b"first")
+    second.write_bytes(b"second")
+
+    result = analyze_paths(serato, music)
+    group = result["details"]["duplicate_filenames"]["items"][0]
+
+    assert len(group["file_previews"]) == 2
+    assert group["file_previews"][0]["audio_url"].startswith("/api/audio?")
+    assert group["file_previews"][0]["reveal_token"]
+
+
+def test_batch_preview_combines_approved_groups_without_changes(tmp_path):
+    serato = tmp_path / "_Serato_"
+    serato.mkdir()
+    music = tmp_path / "Music"
+    choices = []
+    for filename in ("First.mp3", "Second.mp3"):
+        keeper = music / "A" / filename
+        duplicate = music / "B" / filename
+        keeper.parent.mkdir(parents=True, exist_ok=True)
+        duplicate.parent.mkdir(parents=True, exist_ok=True)
+        keeper.write_bytes(b"keeper")
+        duplicate.write_bytes(b"duplicate")
+        choices.append(
+            {"keeper": str(keeper), "group_files": [str(keeper), str(duplicate)]}
+        )
+
+    result = duplicate_repair_batch(
+        serato, music, choices, backup_limit=10
+    )
+
+    assert result["applied"] is False
+    assert result["choice_count"] == 2
+    assert len(result["replaced"]) == 2
+    assert len(result["decisions"]) == 2
+    assert result["decisions"][0]["keeper"].endswith("First.mp3")
+    assert result["decisions"][0]["hash_status"] == "different"
+    assert len(result["decisions"][0]["replaced"]) == 1
+    assert all(not Path(choice["group_files"][1]).is_symlink() for choice in choices)