feat(media): images and recordings in backups, exports and imports (M6)

- Backup ZIP (download and automatic per-user server backups) carries the
  files of the user's own notes as media/<id><ext>; restoring creates
  them with new IDs and rewrites the blocks.
- Markdown archive puts them under attachments/ and embeds them relative
  to each note (![caption](../attachments/…)); re-importing turns them
  back into image/audio blocks.
- Obsidian import takes images and audio a note embeds on its own
  (![[file.png]], ![](path/file.png)); unused or invalid files appear as
  skipped in the preview instead of vanishing.
- Format detection and file paths moved into storage for this.

Co-Authored-By: Claude Opus 5.5 <noreply@anthropic.com>

internal/storage/media.go

@@ -3,6 +3,10 @@ package storage
 import (
 	"database/sql"
 	"errors"
+	"net/http"
+	"path/filepath"
+	"regexp"
+	"strings"
 	"time"
 
 	"github.com/sandrodach/mindsift/internal/domain"
@@ -26,6 +30,72 @@ type Media struct {
 
 var ErrMediaNotFound = errors.New("Datei nicht gefunden")
 
+// Accepted formats (detected from the content, not from the name): file extension and kind
+var mediaTypes = map[string]struct{ ext, kind string }{
+	"image/jpeg":      {".jpg", "image"},
+	"image/png":       {".png", "image"},
+	"image/gif":       {".gif", "image"},
+	"image/webp":      {".webp", "image"},
+	"audio/webm":      {".webm", "audio"},
+	"video/webm":      {".webm", "audio"}, // Chrome's audio-only recordings are detected as video/webm
+	"audio/ogg":       {".ogg", "audio"},
+	"application/ogg": {".ogg", "audio"},
+	"audio/mpeg":      {".mp3", "audio"},
+	"audio/wave":      {".wav", "audio"},
+	"audio/wav":       {".wav", "audio"},
+	"audio/mp4":       {".m4a", "audio"},
+	"video/mp4":       {".m4a", "audio"}, // Safari records MP4/AAC
+	"audio/aac":       {".aac", "audio"},
+}
+
+// DetectMedia recognises the format from the first bytes. Browsers send audio as e.g.
+// "audio/webm;codecs=opus"; that name is kept when the content looks like a container we accept.
+func DetectMedia(data []byte, declared string) (mime, kind string, ok bool) {
+	sniffed := http.DetectContentType(data)
+	if i := strings.IndexByte(sniffed, ';'); i >= 0 {
+		sniffed = sniffed[:i]
+	}
+	// Go does not know WebP and M4A by name; check their signatures directly
+	switch {
+	case len(data) >= 12 && string(data[0:4]) == "RIFF" && string(data[8:12]) == "WEBP":
+		sniffed = "image/webp"
+	case len(data) >= 12 && string(data[4:8]) == "ftyp":
+		sniffed = "video/mp4"
+	}
+	t, ok := mediaTypes[sniffed]
+	if !ok {
+		return "", "", false
+	}
+	if t.kind == "audio" {
+		if d := strings.ToLower(strings.TrimSpace(strings.SplitN(declared, ";", 2)[0])); strings.HasPrefix(d, "audio/") {
+			if dt, known := mediaTypes[d]; known && dt.kind == "audio" {
+				sniffed = d
+			}
+		} else if sniffed == "video/webm" {
+			sniffed = "audio/webm"
+		} else if sniffed == "video/mp4" {
+			sniffed = "audio/mp4"
+		}
+	}
+	return sniffed, t.kind, true
+}
+
+// MediaExt is the file extension for a stored format.
+func MediaExt(mime string) string {
+	if t, ok := mediaTypes[mime]; ok {
+		return t.ext
+	}
+	return ".bin"
+}
+
+// MediaRoot is the folder with all media files ("" for in-memory databases).
+func (s *Store) MediaRoot() string { return s.mediaRoot }
+
+// MediaFilePath is where the file of m lies.
+func (s *Store) MediaFilePath(m *Media) string {
+	return filepath.Join(s.mediaRoot, filepath.Base(m.DocID), filepath.Base(m.ID)+MediaExt(m.Mime))
+}
+
 func (s *Store) migrateMedia() error {
 	return s.execInTx(
 		stmt(`CREATE TABLE IF NOT EXISTS media (
@@ -44,7 +114,11 @@ func (s *Store) migrateMedia() error {
 }
 
 func (s *Store) CreateMedia(docID, ownerID, kind, mime string, size int64, sourceURL string) (*Media, error) {
-	m := &Media{ID: domain.NewID("med"), DocID: docID, OwnerID: ownerID, Kind: kind, Mime: mime, Size: size,
+	return s.createMediaWithID(domain.NewID("med"), docID, ownerID, kind, mime, size, sourceURL)
+}
+
+func (s *Store) createMediaWithID(id, docID, ownerID, kind, mime string, size int64, sourceURL string) (*Media, error) {
+	m := &Media{ID: id, DocID: docID, OwnerID: ownerID, Kind: kind, Mime: mime, Size: size,
 		SourceURL: sourceURL, CreatedAt: nowUTC()}
 	_, err := s.db.Exec(`INSERT INTO media (id, doc_id, owner_id, kind, mime, size, source_url, created_at) VALUES (?, ?, ?, ?, ?, ?, ?, ?)`,
 		m.ID, m.DocID, m.OwnerID, m.Kind, m.Mime, m.Size, m.SourceURL, m.CreatedAt)
@@ -117,3 +191,47 @@ func (s *Store) UnreferencedMedia(olderThan time.Time) ([]Media, error) {
 	}
 	return out, rows.Err()
 }
+
+// MediaOfUser lists the files shown in the user's own notes (for backups and exports).
+func (s *Store) MediaOfUser(userID string) ([]Media, error) {
+	rows, err := s.db.Query(`SELECT DISTINCT m.id, m.doc_id, m.owner_id, m.kind, m.mime, m.size, m.source_url, m.created_at
+		FROM media m JOIN blocks b ON instr(b.content, 'media:' || m.id) > 0
+		JOIN documents d ON d.id = b.document_id WHERE d.user_id = ?`, userID)
+	if err != nil {
+		return nil, err
+	}
+	defer rows.Close()
+	var out []Media
+	for rows.Next() {
+		var m Media
+		if rows.Scan(&m.ID, &m.DocID, &m.OwnerID, &m.Kind, &m.Mime, &m.Size, &m.SourceURL, &m.CreatedAt) == nil {
+			out = append(out, m)
+		}
+	}
+	return out, rows.Err()
+}
+
+// MediaRef is a block's pointer to a file: ![caption](media:id) for images, [caption](media:id) for audio.
+type MediaRef struct{ ID, Caption string }
+
+var mediaRefRe = regexp.MustCompile(`^!?\[((?:[^\]\\]|\\.)*)\]\(media:(med_[0-9a-f]+)\)$`)
+var mediaRefEsc = regexp.MustCompile(`\\(.)`)
+
+func ParseMediaContent(content string) (MediaRef, bool) {
+	m := mediaRefRe.FindStringSubmatch(strings.TrimSpace(content))
+	if m == nil {
+		return MediaRef{}, false
+	}
+	return MediaRef{ID: m[2], Caption: mediaRefEsc.ReplaceAllString(m[1], "$1")}, true
+}
+
+// MediaContent builds a block's content for a file (same format as the browser, lib/media.ts).
+func MediaContent(kind, id, caption string) string {
+	caption = strings.Join(strings.Fields(strings.ReplaceAll(caption, "\n", " ")), " ")
+	caption = strings.NewReplacer(`\`, `\\`, `]`, `\]`).Replace(caption)
+	prefix := ""
+	if kind == "image" {
+		prefix = "!"
+	}
+	return prefix + "[" + caption + "](media:" + id + ")"
+}