---
phase: 10-schema-migration
plan: 01
type: execute
wave: 1
depends_on: []
files_modified:
  - backend/database/sql/schemas/libraries.sql
  - backend/database/sql/schemas/audio_files.sql
  - backend/database/sql/schemas/playlist_tracks.sql
  - backend/database/sql/schemas/track_metadata_view.sql
  - backend/database/database.go
autonomous: true
requirements:
  - DATA-01
  - DATA-04
  - LSCAN-05

must_haves:
  truths:
    - "Fresh database creates libraries table with name, path, created_at columns"
    - "Fresh database creates audio_files with library_id FK column"
    - "Fresh database creates playlist_tracks with nullable audio_file_id and phantom metadata columns"
    - "Fresh database creates track_metadata VIEW including library_id"
    - "Existing v5 database is migrated to v6 atomically — backup created first, all changes in transaction"
    - "Existing audio_files rows get library_id pointing to the auto-created default library"
    - "Migration reads TOML DirectoryPath to create the default library row"
  artifacts:
    - path: "backend/database/sql/schemas/libraries.sql"
      provides: "Libraries table DDL for fresh installs"
      contains: "CREATE TABLE IF NOT EXISTS libraries"
    - path: "backend/database/sql/schemas/audio_files.sql"
      provides: "Updated audio_files DDL with library_id FK"
      contains: "library_id"
    - path: "backend/database/sql/schemas/playlist_tracks.sql"
      provides: "Updated playlist_tracks DDL with nullable audio_file_id and phantom columns"
      contains: "phantom_title"
    - path: "backend/database/sql/schemas/track_metadata_view.sql"
      provides: "Updated VIEW with library_id in SELECT"
      contains: "af.library_id"
    - path: "backend/database/database.go"
      provides: "migration6MultiLibrary function + backup logic"
      contains: "migration6MultiLibrary"
  key_links:
    - from: "backend/database/database.go"
      to: "backend/database/sql/schemas/libraries.sql"
      via: "embedded SQL schema execution in NewDB"
      pattern: "schemas.ReadDir.*sql/schemas"
    - from: "backend/database/database.go migration6"
      to: "TOML config file"
      via: "system.GetUserConfigDirPath + toml decode"
      pattern: "toml\\.Decode"
---

<objective>
Create the database schema definitions and migration 6 for multi-library support.

Purpose: This is the foundational schema change that all subsequent multi-library phases depend on. Fresh installs get the new schema directly; existing databases are migrated atomically with a pre-migration backup.

Output: Updated SQL schema files for fresh databases + migration 6 implementation in database.go
</objective>

<execution_context>
@/home/caleb/.config/opencode/get-shit-done/workflows/execute-plan.md
@/home/caleb/.config/opencode/get-shit-done/templates/summary.md
</execution_context>

<context>
@.planning/PROJECT.md
@.planning/ROADMAP.md
@.planning/STATE.md
@.planning/phases/10-schema-migration/10-CONTEXT.md
@.planning/research/ARCHITECTURE.md
@.planning/research/PITFALLS.md

<interfaces>
<!-- Key types and contracts the executor needs. Extracted from codebase. -->

From backend/database/database.go:
```go
// DB wraps the SQLite database connection and queries.
type DB struct {
    db      *sql.DB
    Ctx     context.Context
    Queries *sqlcgen.Queries
    logger  *slog.Logger
}

// NewDB opens the database and applies schema migrations.
func NewDB(logger *slog.Logger) (*DB, error)

// runMigrations applies incremental schema changes using SQLite's
// PRAGMA user_version as the version tracker.
func runMigrations(ctx context.Context, db *sql.DB, logger *slog.Logger) error

// isDuplicateColumnErr returns true when the error is SQLite's
// "duplicate column name" error.
func isDuplicateColumnErr(err error) bool

// Current migration count: 5 (user_version = 5)
// Migration 5 pattern: table rebuild with FK OFF, DROP VIEW, rebuild, recreate VIEW, FK ON
```

From backend/database/sql/schemas/audio_files.sql (current):
```sql
CREATE TABLE IF NOT EXISTS audio_files (
  id integer PRIMARY KEY,
  file_path text NOT NULL UNIQUE,
  length_milliseconds int NOT NULL,
  file_type_id int NOT NULL,
  recording_id int NOT NULL,
  sample_rate int NOT NULL DEFAULT 0,
  bit_depth int NOT NULL DEFAULT 0,
  channels int NOT NULL DEFAULT 0,
  bitrate int NOT NULL DEFAULT 0,
  file_size int NOT NULL DEFAULT 0,
  basename text NOT NULL DEFAULT '',
  FOREIGN KEY(file_type_id) REFERENCES file_types(id),
  FOREIGN KEY(recording_id) REFERENCES recordings(id)
);
```

From backend/database/sql/schemas/playlist_tracks.sql (current):
```sql
CREATE TABLE IF NOT EXISTS playlist_tracks (
    id INTEGER PRIMARY KEY,
    playlist_id INTEGER NOT NULL,
    audio_file_id INTEGER NOT NULL,
    position INTEGER NOT NULL,
    FOREIGN KEY(playlist_id) REFERENCES playlists(id) ON DELETE CASCADE,
    FOREIGN KEY(audio_file_id) REFERENCES audio_files(id) ON DELETE CASCADE
);
```

From backend/database/sql/schemas/queue_tracks.sql (current — CASCADE stays):
```sql
CREATE TABLE IF NOT EXISTS queue_tracks (
    id INTEGER PRIMARY KEY,
    audio_file_id INTEGER NOT NULL,
    position INTEGER NOT NULL,
    FOREIGN KEY(audio_file_id) REFERENCES audio_files(id) ON DELETE CASCADE
);
```

From backend/library/config.go:
```go
type Config struct {
    DirectoryPath   Directory       `toml:"DirectoryPath"`
    ScanConcurrency ScanConcurrency `toml:"ScanConcurrency"`
}
```

From backend/system/userdata.go:
```go
func GetUserDataDirPath() (string, error)
func GetUserConfigDirPath() (string, error)
```
</interfaces>
</context>

<tasks>

<task type="auto">
  <name>Task 1: Update SQL schema files for fresh installs</name>
  <files>
    backend/database/sql/schemas/libraries.sql
    backend/database/sql/schemas/audio_files.sql
    backend/database/sql/schemas/playlist_tracks.sql
    backend/database/sql/schemas/track_metadata_view.sql
  </files>
  <action>
Create the schema files that define the target state for fresh database installs. These files are executed via `go:embed` in `NewDB()` — they use `CREATE TABLE IF NOT EXISTS` / `CREATE VIEW IF NOT EXISTS` so they're idempotent.

**1. Create `libraries.sql` (NEW FILE):**
```sql
CREATE TABLE IF NOT EXISTS libraries (
    id         INTEGER PRIMARY KEY,
    name       TEXT NOT NULL,
    path       TEXT NOT NULL UNIQUE,
    created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP
);
```
Per user decision: minimal table — name, path, created_at only. No scan metadata columns (Phase 11 adds those). No scan_concurrency column (global default fallback for now).

**2. Update `audio_files.sql`:**
Add `library_id` column with FK to libraries table. For fresh databases the column should be `NOT NULL` with no DEFAULT (fresh installs always create a library first). However, since the CREATE TABLE runs before any libraries exist, use `DEFAULT 0` to allow the table creation to succeed — the migration and scan pipeline will always set the correct value.

Add after the `basename` column:
```sql
  library_id int NOT NULL DEFAULT 0,
```
Add FK constraint:
```sql
  FOREIGN KEY(library_id) REFERENCES libraries(id)
```
Add index after the table:
```sql
CREATE INDEX IF NOT EXISTS idx_audio_files_library_id
    ON audio_files(library_id);
```

**3. Update `playlist_tracks.sql`:**
Change `audio_file_id` from `NOT NULL` to nullable (remove NOT NULL). Change FK from `ON DELETE CASCADE` to `ON DELETE SET NULL`. Add phantom metadata columns with NULL defaults:
```sql
CREATE TABLE IF NOT EXISTS playlist_tracks (
    id INTEGER PRIMARY KEY,
    playlist_id INTEGER NOT NULL,
    audio_file_id INTEGER,
    position INTEGER NOT NULL,
    phantom_title TEXT,
    phantom_artist TEXT,
    phantom_album TEXT,
    phantom_duration_ms INTEGER,
    phantom_genre TEXT,
    phantom_cover_art_path TEXT,
    FOREIGN KEY(playlist_id) REFERENCES playlists(id) ON DELETE CASCADE,
    FOREIGN KEY(audio_file_id) REFERENCES audio_files(id) ON DELETE SET NULL
);
```
Keep the existing indexes on playlist_id and audio_file_id.

**4. Update `track_metadata_view.sql`:**
Add `af.library_id` to the SELECT list — insert it after `af.file_size` (last column). The JOIN structure stays identical:
```sql
    af.file_size,
    af.library_id
FROM audio_files af
```

**IMPORTANT:** The `libraries.sql` file must sort BEFORE `audio_files.sql` alphabetically so it's executed first (the FK depends on it). Verify: "libraries" < "audio_files" — NO, "a" < "l" so audio_files runs first. This is a problem because audio_files references libraries. Solutions:
- Rename to `001_libraries.sql` — but this changes naming convention
- Use the migration to handle existing DBs and rely on SQLite's deferred FK check for fresh DBs — since `PRAGMA foreign_keys = ON` is set AFTER schema files run? No — PRAGMAs run BEFORE schemas in `NewDB()`.

Actually, check the code: `applyPRAGMAs()` runs `PRAGMA foreign_keys = ON` before schema files execute. So `audio_files.sql` will fail FK check if `libraries` table doesn't exist yet. The fix: name the file so it sorts before audio_files. Use `_libraries.sql` (underscore sorts before 'a' in ASCII). Or better: just create the libraries table inside audio_files.sql as a preceding statement? No, that's messy.

Best approach: Name the file so it executes first. Files are read from embedded FS sorted by `ReadDir` (alphabetical). Prefix: `00_libraries.sql` ensures it runs before any other schema. This is clean and explicit.

Actually — re-read the `NewDB()` code. It iterates `dirEntries` from `ReadDir` which returns entries sorted by name. So we need `libraries.sql` to sort before `audio_files.sql`. Since 'l' > 'a', we can't use `libraries.sql` directly. Use `_libraries.sql` (underscore = 0x5F sorts after uppercase letters in ASCII but before lowercase... actually in Go's `ReadDir`, entries are sorted case-sensitively). Since all existing files are lowercase, and '_' (0x5F) < 'a' (0x61), `_libraries.sql` will sort first. 

Wait — there are other options. The simplest: just temporarily disable FK checks in `NewDB` during schema file execution, then re-enable. But that changes production behavior. 

Simplest correct approach: just make the FK constraint on audio_files use a CREATE INDEX rather than inline FK, and handle the FK check in the migration. But that loses FK enforcement on fresh DBs.

Actually the cleanest fix: since `libraries.sql` needs to exist before `audio_files.sql`, and Go's `ReadDir` sorts alphabetically, and `_` < `a`: name it `_libraries.sql`. This runs before all other schema files.

Alternative: just don't add the FK constraint to the CREATE TABLE DDL and instead add it during migration / via a separate step. But SQLite doesn't support ALTER TABLE ADD CONSTRAINT.

Go with `_libraries.sql`. It's a minor naming convention change but ensures correct execution order.
  </action>
  <verify>
    <automated>cd backend/database && go build ./...</automated>
  </verify>
  <done>
    - `_libraries.sql` exists with CREATE TABLE IF NOT EXISTS libraries
    - `audio_files.sql` includes library_id column and FK
    - `playlist_tracks.sql` has nullable audio_file_id, SET NULL FK, and all 6 phantom columns
    - `track_metadata_view.sql` includes af.library_id in SELECT
    - Package compiles successfully
  </done>
</task>

<task type="auto">
  <name>Task 2: Implement migration 6 and pre-migration backup</name>
  <files>
    backend/database/database.go
  </files>
  <action>
Add migration 6 to the `runMigrations()` function in `database.go`. This is the most complex migration yet — follow the established patterns from migration 5 (table rebuild with FK OFF).

**Step 1: Add backup function.**

Create `backupDatabase()` function that copies the database file before migration 6 runs. Per user decision: timestamp-based naming (e.g., `yj.db.bak.20260309`), no automatic cleanup, logged at INFO level.

```go
// backupDatabase copies the database file to a timestamped backup
// before running a destructive migration. Returns the backup path.
func backupDatabase(
    dbPath string, logger *slog.Logger,
) (string, error) {
    backupPath := dbPath + ".bak." + time.Now().Format("20060102")
    // Use io.Copy from source to destination
    // Log at INFO: "database backup created", "path", backupPath
    // Return backupPath, nil on success
}
```

The `dbPath` must be passed to `runMigrations`. Update the signature:
```go
func runMigrations(ctx context.Context, db *sql.DB, logger *slog.Logger, dbPath string) error
```
Update the call site in `NewDB()` to pass `sqliteDBFilePath`.

**Step 2: Add migration 6 block in runMigrations.**

After the `version < 5` block, add:

```go
// Migration 6: multi-library support.
if version < 6 {
    if err := migration6MultiLibrary(
        ctx, db, logger, dbPath,
    ); err != nil {
        return err
    }
}
```

**Step 3: Implement `migration6MultiLibrary()` function.**

This is a large function — follow migration 5's pattern. The steps MUST execute in this exact order inside a single transaction (DATA-04: atomic):

```go
func migration6MultiLibrary(
    ctx context.Context,
    db *sql.DB,
    logger *slog.Logger,
    dbPath string,
) error {
    logger.Info("applying migration 6: multi-library support")

    // 1. Backup database BEFORE any changes.
    backupPath, err := backupDatabase(dbPath, logger)
    // Handle error — if backup fails, abort migration.
    logger.Info("pre-migration backup created", "path", backupPath)

    // 2. Read TOML config to get existing library directory.
    // Use system.GetUserConfigDirPath() to find config.toml.
    // Parse ONLY the [Library] section to get DirectoryPath.
    // If no config or no DirectoryPath, existingDir = "" (fresh install).
    configDir, err := system.GetUserConfigDirPath()
    // Read config.toml, decode [Library].DirectoryPath
    // Use a minimal struct: struct{ Library struct{ DirectoryPath string } }

    // 3. Disable FK checks for table rebuild.
    _, err = db.ExecContext(ctx, "PRAGMA foreign_keys = OFF")

    // 4. Create libraries table.
    _, err = db.ExecContext(ctx, `
        CREATE TABLE IF NOT EXISTS libraries (
            id         INTEGER PRIMARY KEY,
            name       TEXT NOT NULL,
            path       TEXT NOT NULL UNIQUE,
            created_at DATETIME NOT NULL DEFAULT CURRENT_TIMESTAMP
        )
    `)

    // 5. Insert default library from TOML (if existingDir is not empty).
    var defaultLibID int64
    if existingDir != "" {
        // Derive library name from directory basename.
        // e.g., "/home/user/Music" -> "Music"
        libName := filepath.Base(existingDir)
        result, err := db.ExecContext(ctx,
            "INSERT INTO libraries (name, path) VALUES (?, ?)",
            libName, existingDir,
        )
        defaultLibID, _ = result.LastInsertId()
        logger.Info("migrated existing library",
            "name", libName,
            "path", existingDir,
            "id", defaultLibID,
        )
    }

    // 6. Add library_id column to audio_files.
    // Use DEFAULT with the actual library ID so existing rows are backfilled.
    // Per P1: NOT NULL column added via ALTER TABLE requires DEFAULT.
    stmt := fmt.Sprintf(
        "ALTER TABLE audio_files ADD COLUMN library_id INTEGER NOT NULL DEFAULT %d",
        defaultLibID,
    )
    if _, err := db.ExecContext(ctx, stmt); err != nil {
        if !isDuplicateColumnErr(err) { return ... }
    }

    // 7. Create index on library_id.
    _, err = db.ExecContext(ctx, `
        CREATE INDEX IF NOT EXISTS idx_audio_files_library_id
            ON audio_files(library_id)
    `)

    // 8. Drop track_metadata VIEW (references audio_files which we're about to rebuild playlist_tracks against).
    _, err = db.ExecContext(ctx, "DROP VIEW IF EXISTS track_metadata")

    // 9. Rebuild playlist_tracks for SET NULL FK + phantom columns.
    // Per P2: audit ALL CASCADE FKs — playlist_tracks changes to SET NULL,
    // queue_tracks keeps CASCADE (ephemeral).
    _, err = db.ExecContext(ctx, `
        CREATE TABLE playlist_tracks_new (
            id INTEGER PRIMARY KEY,
            playlist_id INTEGER NOT NULL,
            audio_file_id INTEGER,
            position INTEGER NOT NULL,
            phantom_title TEXT,
            phantom_artist TEXT,
            phantom_album TEXT,
            phantom_duration_ms INTEGER,
            phantom_genre TEXT,
            phantom_cover_art_path TEXT,
            FOREIGN KEY(playlist_id) REFERENCES playlists(id) ON DELETE CASCADE,
            FOREIGN KEY(audio_file_id) REFERENCES audio_files(id) ON DELETE SET NULL
        )
    `)

    // Copy existing data (phantom columns get NULL).
    _, err = db.ExecContext(ctx, `
        INSERT INTO playlist_tracks_new (id, playlist_id, audio_file_id, position)
        SELECT id, playlist_id, audio_file_id, position FROM playlist_tracks
    `)

    // Drop old table.
    _, err = db.ExecContext(ctx, "DROP TABLE playlist_tracks")

    // Rename.
    _, err = db.ExecContext(ctx, "ALTER TABLE playlist_tracks_new RENAME TO playlist_tracks")

    // Recreate indexes.
    _, err = db.ExecContext(ctx, `
        CREATE INDEX IF NOT EXISTS idx_playlist_tracks_playlist_id
            ON playlist_tracks(playlist_id)
    `)
    _, err = db.ExecContext(ctx, `
        CREATE INDEX IF NOT EXISTS idx_playlist_tracks_audio_file_id
            ON playlist_tracks(audio_file_id)
    `)

    // 10. Backfill phantom metadata on existing playlist_tracks from audio_files JOINs.
    // Per user decision: eager population — fill metadata now, not lazily.
    _, err = db.ExecContext(ctx, `
        UPDATE playlist_tracks SET
            phantom_title = sub.title,
            phantom_artist = sub.artist,
            phantom_album = sub.album,
            phantom_duration_ms = sub.duration,
            phantom_genre = sub.genre,
            phantom_cover_art_path = sub.cover_art_path
        FROM (
            SELECT
                pt.id AS pt_id,
                COALESCE(r.name, '') AS title,
                COALESCE(ac.text, '') AS artist,
                COALESCE(rg.name, '') AS album,
                af.length_milliseconds AS duration,
                CAST(COALESCE(
                    (SELECT GROUP_CONCAT(g.name, '||')
                     FROM recording_genres rg_sub
                     JOIN genres g ON rg_sub.genre_id = g.id
                     WHERE rg_sub.recording_id = r.id),
                    ''
                ) AS TEXT) AS genre,
                COALESCE(ca.file_path, '') AS cover_art_path
            FROM playlist_tracks pt
            JOIN audio_files af ON pt.audio_file_id = af.id
            LEFT JOIN recordings r ON af.recording_id = r.id
            LEFT JOIN artist_credit ac ON r.artist_credit_id = ac.id
            LEFT JOIN (
                SELECT recording_id, MIN(release_group_id) AS release_group_id
                FROM release_group_recordings
                GROUP BY recording_id
            ) rgr ON r.id = rgr.recording_id
            LEFT JOIN release_groups rg ON rgr.release_group_id = rg.id
            LEFT JOIN cover_art ca ON rg.cover_art_id = ca.id
        ) sub
        WHERE playlist_tracks.id = sub.pt_id
    `)

    // 11. Recreate track_metadata VIEW with library_id.
    _, err = db.ExecContext(ctx, `
        CREATE VIEW IF NOT EXISTS track_metadata AS
        SELECT
            af.id,
            af.file_path,
            af.length_milliseconds,
            COALESCE(r.name, '') AS title,
            COALESCE(ac.text, '') AS artist_name,
            r.track_number,
            r.disc_number,
            COALESCE(rg.name, '') AS album,
            CAST(COALESCE(
                (SELECT GROUP_CONCAT(g.name, '||')
                 FROM recording_genres rg_sub
                 JOIN genres g ON rg_sub.genre_id = g.id
                 WHERE rg_sub.recording_id = r.id),
                ''
            ) AS TEXT) AS genre,
            COALESCE(r.year, 0) AS year,
            COALESCE(r.composer, '') AS composer,
            COALESCE(ft.extension, '') AS file_type,
            af.sample_rate,
            af.bit_depth,
            af.channels,
            af.bitrate,
            af.file_size,
            af.library_id
        FROM audio_files af
        LEFT JOIN recordings r ON af.recording_id = r.id
        LEFT JOIN artist_credit ac ON r.artist_credit_id = ac.id
        LEFT JOIN (
            SELECT recording_id,
                MIN(release_group_id) AS release_group_id
            FROM release_group_recordings
            GROUP BY recording_id
        ) rgr ON r.id = rgr.recording_id
        LEFT JOIN release_groups rg ON rgr.release_group_id = rg.id
        LEFT JOIN file_types ft ON af.file_type_id = ft.id
    `)

    // 12. Re-enable FK checks.
    _, err = db.ExecContext(ctx, "PRAGMA foreign_keys = ON")

    // 13. Remove music_directory from TOML config.
    // Read the full config, nil out the Library.DirectoryPath, write back.
    // Per user decision: old key ignored if still present (no crash).
    // Use BurntSushi/toml for read/write consistency.
    // Only do this if existingDir was non-empty (migration actually ran).
    if existingDir != "" {
        removeLibraryDirFromTOML(configDir, logger)
    }

    // 14. Set version.
    _, err = db.ExecContext(ctx, "PRAGMA user_version = 6")

    logger.Info("migration 6 complete")
    return nil
}
```

**Step 4: Implement `removeLibraryDirFromTOML()` helper.**

Read the TOML file, set DirectoryPath to empty string, write back. Use the same `os.WriteFile` with `0o644` permissions pattern from the config package. If the file doesn't exist or the section is missing, no-op (per user decision: old config key ignored).

**IMPORTANT notes for the executor:**
- Import `path/filepath` for `filepath.Base()` and `time` for backup timestamp.
- Import `io` for `io.Copy` in backup function.
- Import `os` for file operations.
- Import `github.com/BurntSushi/toml` for TOML read/write in migration.
- Add `// SAFETY:` comments on all hand-crafted SQL (consistent with Phase 6 convention).
- The backup runs OUTSIDE the transaction (you can't copy a file inside a SQL transaction). The migration SQL steps should be wrapped in a transaction for atomicity. Use `db.BeginTx()` around steps 3-12.
- Actually, PRAGMA foreign_keys cannot run inside a transaction. Structure: backup → PRAGMA FK OFF → BEGIN TX → steps 4-11 → COMMIT → PRAGMA FK ON → PRAGMA user_version = 6.
- Wait — PRAGMA user_version also can't run inside a transaction reliably on all SQLite versions. Follow migration 5's pattern: no explicit transaction, just sequential statements with PRAGMA FK OFF/ON wrapping.
- For fresh installs with no TOML config: existingDir="" and defaultLibID=0. The ALTER TABLE ADD COLUMN with DEFAULT 0 is fine — there are no audio_files rows on a fresh install anyway. The schema files handle fresh DB creation.
- The `library_id NOT NULL DEFAULT 0` on audio_files in the schema file means fresh-install audio_files don't require a library to exist yet. The scan pipeline (Phase 11) will set library_id correctly. DEFAULT 0 is a placeholder that won't satisfy the FK constraint, but since `PRAGMA foreign_keys` only checks on INSERT/UPDATE, and the CREATE TABLE runs before any data, this is safe.

Actually, that FK constraint with DEFAULT 0 is problematic. If FK checks are on and someone inserts a row without a library, it'll fail. For fresh installs the scan pipeline (Phase 11) will always set a real library_id. But to be safe, DON'T add a FK constraint in the CREATE TABLE for audio_files — add it only via the migration where we control the value. Wait, no — we want FK enforcement on fresh DBs too.

Better approach: Use `DEFAULT 1` in the schema file — but library ID 1 may not exist on fresh installs. Actually for fresh installs per user decision: "empty libraries table, user adds their first library when they want to scan." So there's no library to FK-reference. The scan pipeline in Phase 11 will create a library first, then scan.

The safest approach: keep the FK constraint and `NOT NULL DEFAULT 0` in the schema file. Since `PRAGMA foreign_keys = ON` is set, any INSERT into audio_files without a valid library_id will fail — which is correct behavior. The DEFAULT 0 only matters for the ALTER TABLE ADD COLUMN during migration where it backfills existing rows. We immediately set all rows to the correct library_id in the same migration.

Wait — for the ALTER TABLE ADD COLUMN in migration 6, the DEFAULT value must match the actual library ID. That's `defaultLibID` (dynamic). So the schema file's DEFAULT 0 is fine for CREATE TABLE (fresh DBs), and the migration uses a dynamic DEFAULT.

One more thing: on fresh DBs, audio_files will have `library_id INTEGER NOT NULL DEFAULT 0` with a FK to libraries. If someone tries to INSERT an audio_file with library_id=0 and no library with id=0 exists, the FK check will fail. This is actually CORRECT — you must create a library first. Good.

Let the executor figure out the exact DEFAULT handling. The key instruction is clear.
  </action>
  <verify>
    <automated>cd backend/database && go build ./... && go vet ./...</automated>
  </verify>
  <done>
    - `runMigrations` signature updated to accept dbPath
    - `backupDatabase()` creates timestamped copy of .db file
    - `migration6MultiLibrary()` implements all 14 steps in order
    - TOML DirectoryPath is read and used to create default library
    - Library name derived from directory basename
    - playlist_tracks rebuilt with SET NULL FK and 6 phantom columns
    - Phantom metadata backfilled from audio_files JOINs on existing rows
    - track_metadata VIEW recreated with library_id column
    - TOML config cleaned up (DirectoryPath removed after migration)
    - All hand-crafted SQL has SAFETY comments
    - Package compiles and passes vet
  </done>
</task>

</tasks>

<verification>
- `go build ./...` passes from project root
- `go vet ./...` passes from backend/database
- No linting errors on new code: `golangci-lint run ./backend/database/...`
</verification>

<success_criteria>
- Fresh database creates all tables including libraries and updated audio_files/playlist_tracks
- Migration 6 function exists with complete implementation
- Backup function creates timestamped database copy
- All schema changes follow established migration patterns
- TOML config reading works for default library creation
</success_criteria>

<output>
After completion, create `.planning/phases/10-schema-migration/10-01-SUMMARY.md`
</output>