Make photo library offline-first and fix video thumbnails

Server/services/thumbnails/pkg/preprocessor/preprocessor.go

@@ -4,15 +4,20 @@ import (
 	"archive/zip"
 	"bufio"
 	"bytes"
+	"context"
 	"encoding/base64"
 	"encoding/json"
 	"image"
 	"image/draw"
 	"image/gif"
+	_ "image/jpeg"
 	"io"
 	"math"
 	"mime"
+	"os"
+	"os/exec"
 	"strings"
+	"time"
 
 	"github.com/pkg/errors"
 	"golang.org/x/image/font"
@@ -41,6 +46,59 @@ func (i GifDecoder) Convert(r io.Reader) (any, error) {
 	return img, nil
 }
 
+// VideoDecoder extracts a bounded preview frame with ffmpeg. The input is first persisted to a
+// temporary file so ffmpeg can seek to MP4/MOV metadata stored at the end of large camera files.
+// The file lives on disk and is never retained in process memory.
+type VideoDecoder struct{}
+
+func (VideoDecoder) Convert(r io.Reader) (any, error) {
+	input, err := os.CreateTemp("", "qsfera-video-*")
+	if err != nil {
+		return nil, err
+	}
+	inputPath := input.Name()
+	defer os.Remove(inputPath)
+	if _, err = io.Copy(input, r); err != nil {
+		input.Close()
+		return nil, errors.Wrap(err, "could not persist video for preview extraction")
+	}
+	if err = input.Sync(); err != nil {
+		input.Close()
+		return nil, err
+	}
+	if err = input.Close(); err != nil {
+		return nil, err
+	}
+
+	ctx, cancel := context.WithTimeout(context.Background(), videoFrameExtractionTimeout)
+	defer cancel()
+	frame, err := extractVideoFrame(ctx, inputPath, nil)
+	if err != nil {
+		return nil, errors.Wrap(err, "could not extract video preview frame")
+	}
+
+	decoded, _, err := image.Decode(bytes.NewReader(frame))
+	if err != nil {
+		return nil, errors.Wrap(err, "could not decode video preview frame")
+	}
+	return decoded, nil
+}
+
+const videoFrameExtractionTimeout = 90 * time.Second
+
+func extractVideoFrame(ctx context.Context, input string, stdin io.Reader) ([]byte, error) {
+	cmd := exec.CommandContext(
+		ctx,
+		"ffmpeg",
+		"-hide_banner", "-loglevel", "error", "-i", input,
+		"-map", "0:v:0", "-an", "-sn", "-frames:v", "1",
+		"-vf", "thumbnail=30,scale=1920:-2:force_original_aspect_ratio=decrease",
+		"-f", "image2pipe", "-vcodec", "png", "pipe:1",
+	)
+	cmd.Stdin = stdin
+	return cmd.Output()
+}
+
 // GgsDecoder is a converter for the geogebra slides file
 type GgsDecoder struct{ thumbnailpath string }
 
@@ -303,6 +361,9 @@ func ForType(mimeType string, opts map[string]any) FileConverter {
 	// return the service call. So we should only get here when the mimeType parses fine.
 	mimeType, _, _ = mime.ParseMediaType(mimeType)
 	switch mimeType {
+	case "video/mp4", "video/quicktime", "video/webm", "video/x-matroska", "video/x-msvideo", "video/mpeg", "video/3gpp",
+		"video/x-m4v", "video/mp2t", "video/ogg", "video/x-ms-wmv", "video/x-flv", "video/hevc":
+		return VideoDecoder{}
 	case "text/plain":
 		fontFileMap := ""
 		fontFaceOpts := &opentype.FaceOptions{