add image conversion for telegram

2026-02-13 09:37:54 +03:00
parent 1905e7ab16
commit e1ebb83641
5 changed files with 203 additions and 35 deletions
@@ -0,0 +1,66 @@
+package media
+
+import (
+	"bytes"
+	"context"
+	"fmt"
+	"os/exec"
+	"strings"
+)
+
+const maxTelegramPhotoSize = 10 * 1024 * 1024 // 10 MB
+
+// CompressImage compresses an image if it exceeds Telegram's 10 MB photo limit.
+// It uses ImageMagick's magick command to convert via stdin/stdout.
+// Returns the (possibly compressed) data, updated filename, and any error.
+func CompressImage(ctx context.Context, data []byte, filename string) (out []byte, outName string, err error) {
+	if len(data) <= maxTelegramPhotoSize {
+		return data, filename, nil
+	}
+
+	// First attempt: just re-encode as JPEG with quality 85
+	out, err = runMagick(ctx, data, "-quality", "85")
+	if err != nil {
+		return nil, "", fmt.Errorf("compress image: %w", err)
+	}
+
+	newFilename := replaceExt(filename, ".jpg")
+
+	if len(out) <= maxTelegramPhotoSize {
+		return out, newFilename, nil
+	}
+
+	// Second attempt: resize to 50% and quality 85
+	out, err = runMagick(ctx, data, "-resize", "50%", "-quality", "85")
+	if err != nil {
+		return nil, "", fmt.Errorf("compress image with resize: %w", err)
+	}
+
+	if len(out) > maxTelegramPhotoSize {
+		return nil, "", fmt.Errorf("image still too large after compression (%d bytes)", len(out))
+	}
+
+	return out, newFilename, nil
+}
+
+func runMagick(ctx context.Context, data []byte, args ...string) ([]byte, error) {
+	cmdArgs := append([]string{"-"}, args...)
+	cmdArgs = append(cmdArgs, "jpeg:-")
+
+	cmd := exec.CommandContext(ctx, "magick", cmdArgs...)
+	cmd.Stdin = bytes.NewReader(data)
+
+	out, err := cmd.Output()
+	if err != nil {
+		return nil, fmt.Errorf("magick: %w", err)
+	}
+
+	return out, nil
+}
+
+func replaceExt(filename, newExt string) string {
+	if i := strings.LastIndex(filename, "."); i >= 0 {
+		return filename[:i] + newExt
+	}
+	return filename + newExt
+}
@@ -37,6 +37,24 @@ func (a Attachment) IsImage() bool {
 	return false
 }

+// IsVideo returns true if the attachment is a video.
+func (a Attachment) IsVideo() bool {
+	switch a.Type {
+	case "video/mp4", "video/quicktime", "video/webm", "video/mpeg":
+		return true
+	}
+	return false
+}
+
+// IsAudio returns true if the attachment is an audio file.
+func (a Attachment) IsAudio() bool {
+	switch a.Type {
+	case "audio/mpeg", "audio/ogg", "audio/wav", "audio/mp4", "audio/aac":
+		return true
+	}
+	return false
+}
+
 // UID extracts the uid part from "attachments/{uid}".
 func (a Attachment) UID() string {
 	const prefix = "attachments/"
@@ -11,6 +11,7 @@ import (
 	tgbotapi "github.com/go-telegram-bot-api/telegram-bot-api/v5"

 	"git.vakhrushev.me/av/remembos/internal/config"
+	"git.vakhrushev.me/av/remembos/internal/media"
 	"git.vakhrushev.me/av/remembos/internal/memory"
 	"git.vakhrushev.me/av/remembos/internal/memos"
 	"git.vakhrushev.me/av/remembos/internal/search"
@@ -130,14 +131,22 @@ func (b *Bot) sendMemory(ctx context.Context, mem *search.Memory) {
 	}

 	text := formatMemory(mem, b.publicURL)
-	images := imageAttachments(mem.Memo)
+	imageAtts, videoAtts, audioAtts := mediaAttachments(mem.Memo)

-	var downloaded []imageFile
-	if len(images) > 0 {
-		downloaded = b.downloadImages(ctx, images)
+	var images []mediaFile
+	var skipped bool
+	if len(imageAtts) > 0 {
+		images, skipped = b.downloadAndCompressImages(ctx, imageAtts)
 	}

-	if err := b.sendWithRetry(ctx, text, downloaded); err != nil {
+	videos := b.downloadFiles(ctx, videoAtts)
+	audios := b.downloadFiles(ctx, audioAtts)
+
+	if skipped {
+		text += "\n\nПоказаны не все вложения"
+	}
+
+	if err := b.sendWithRetry(ctx, text, images, videos, audios); err != nil {
 		b.logger.Error("failed to send telegram message after retries", "error", err)
 	}
 }
@@ -183,35 +192,57 @@ func (b *Bot) handleMore(ctx context.Context) {
 	b.sendMemory(ctx, mem)
 }

-type imageFile struct {
+type mediaFile struct {
 	filename string
 	data     []byte
 }

-// downloadImages downloads image attachments, skipping failures.
-func (b *Bot) downloadImages(ctx context.Context, attachments []memos.Attachment) []imageFile {
-	files := make([]imageFile, 0, len(attachments))
+// downloadAndCompressImages downloads image attachments and compresses them if needed.
+// Returns downloaded files and whether any were skipped due to errors.
+func (b *Bot) downloadAndCompressImages(ctx context.Context, attachments []memos.Attachment) ([]mediaFile, bool) {
+	files := make([]mediaFile, 0, len(attachments))
+	var skipped bool
+	for _, att := range attachments {
+		data, err := b.client.DownloadAttachment(ctx, att)
+		if err != nil {
+			b.logger.Warn("failed to download attachment, skipping", "name", att.Name, "error", err)
+			skipped = true
+			continue
+		}
+
+		data, filename, err := media.CompressImage(ctx, data, att.Filename)
+		if err != nil {
+			b.logger.Warn("failed to compress image, skipping", "name", att.Name, "error", err)
+			skipped = true
+			continue
+		}
+
+		files = append(files, mediaFile{filename: filename, data: data})
+	}
+	return files, skipped
+}
+
+// downloadFiles downloads attachments, skipping failures.
+func (b *Bot) downloadFiles(ctx context.Context, attachments []memos.Attachment) []mediaFile {
+	files := make([]mediaFile, 0, len(attachments))
 	for _, att := range attachments {
 		data, err := b.client.DownloadAttachment(ctx, att)
 		if err != nil {
 			b.logger.Warn("failed to download attachment, skipping", "name", att.Name, "error", err)
 			continue
 		}
-		files = append(files, imageFile{
-			filename: att.Filename,
-			data:     data,
-		})
+		files = append(files, mediaFile{filename: att.Filename, data: data})
 	}
 	return files
 }

 // sendWithRetry attempts to send the message with up to 3 retries.
-func (b *Bot) sendWithRetry(ctx context.Context, text string, images []imageFile) error {
+func (b *Bot) sendWithRetry(ctx context.Context, text string, images, videos, audios []mediaFile) error {
 	backoffs := []time.Duration{30 * time.Second, 60 * time.Second, 120 * time.Second}
 	var lastErr error

 	for attempt := range 3 {
-		lastErr = b.send(text, images)
+		lastErr = b.sendWithMedia(text, images, videos, audios)
 		if lastErr == nil {
 			return nil
 		}
@@ -228,31 +259,59 @@ func (b *Bot) sendWithRetry(ctx context.Context, text string, images []imageFile
 	return lastErr
 }

-// send executes the actual Telegram API calls based on the sending strategy.
-// Long text is split into multiple messages. Images are sent with a caption
-// only if the full text fits within the caption limit.
-func (b *Bot) send(text string, images []imageFile) error {
+// sendWithMedia executes the actual Telegram API calls for text, images, videos, and audios.
+func (b *Bot) sendWithMedia(text string, images, videos, audios []mediaFile) error {
+	hasMedia := len(images) > 0 || len(videos) > 0 || len(audios) > 0
+
+	// Send text
 	switch {
-	case len(images) == 0:
+	case !hasMedia:
 		return b.sendTextParts(splitText(text, maxMessageLen))

-	case len(text) <= maxCaptionLen:
+	case len(text) <= maxCaptionLen && len(images) > 0:
 		// Short text — use as caption on image(s)
 		if len(images) == 1 {
-			return b.sendPhoto(images[0], text)
+			if err := b.sendPhoto(images[0], text); err != nil {
+				return err
+			}
+		} else {
+			if err := b.sendMediaGroup(images, text); err != nil {
+				return err
+			}
 		}
-		return b.sendMediaGroup(images, text)

 	default:
-		// Long text — send text messages first, then images without caption
+		// Long text or no images — send text first, then all media
 		if err := b.sendTextParts(splitText(text, maxMessageLen)); err != nil {
 			return err
 		}
+		// Send images without caption
 		if len(images) == 1 {
-			return b.sendPhoto(images[0], "")
+			if err := b.sendPhoto(images[0], ""); err != nil {
+				return err
+			}
+		} else if len(images) > 1 {
+			if err := b.sendMediaGroup(images, ""); err != nil {
+				return err
+			}
 		}
-		return b.sendMediaGroup(images, "")
 	}
+
+	// Send videos one by one
+	for _, v := range videos {
+		if err := b.sendVideo(v); err != nil {
+			return err
+		}
+	}
+
+	// Send audios one by one
+	for _, a := range audios {
+		if err := b.sendAudio(a); err != nil {
+			return err
+		}
+	}
+
+	return nil
 }

 func (b *Bot) sendTextParts(parts []string) error {
@@ -272,7 +331,7 @@ func (b *Bot) sendText(text string) error {
 	return err
 }

-func (b *Bot) sendPhoto(img imageFile, caption string) error {
+func (b *Bot) sendPhoto(img mediaFile, caption string) error {
 	photo := tgbotapi.NewPhoto(b.chatID, tgbotapi.FileBytes{
 		Name:  img.filename,
 		Bytes: img.data,
@@ -285,7 +344,25 @@ func (b *Bot) sendPhoto(img imageFile, caption string) error {
 	return err
 }

-func (b *Bot) sendMediaGroup(images []imageFile, caption string) error {
+func (b *Bot) sendVideo(v mediaFile) error {
+	video := tgbotapi.NewVideo(b.chatID, tgbotapi.FileBytes{
+		Name:  v.filename,
+		Bytes: v.data,
+	})
+	_, err := b.api.Send(video)
+	return err
+}
+
+func (b *Bot) sendAudio(a mediaFile) error {
+	audio := tgbotapi.NewAudio(b.chatID, tgbotapi.FileBytes{
+		Name:  a.filename,
+		Bytes: a.data,
+	})
+	_, err := b.api.Send(audio)
+	return err
+}
+
+func (b *Bot) sendMediaGroup(images []mediaFile, caption string) error {
 	media := make([]interface{}, len(images))
 	for i, img := range images {
 		photo := tgbotapi.NewInputMediaPhoto(tgbotapi.FileBytes{
@@ -46,15 +46,19 @@ func formatMemory(mem *search.Memory, publicURL string) string {
 	return b.String()
 }

-// imageAttachments returns image attachments from the memo.
-func imageAttachments(memo *memos.Memo) []memos.Attachment {
-	var images []memos.Attachment
+// mediaAttachments splits memo attachments into images, videos, and audios.
+func mediaAttachments(memo *memos.Memo) (images, videos, audios []memos.Attachment) {
 	for _, att := range memo.Attachments {
-		if att.IsImage() {
+		switch {
+		case att.IsImage():
 			images = append(images, att)
+		case att.IsVideo():
+			videos = append(videos, att)
+		case att.IsAudio():
+			audios = append(audios, att)
 		}
 	}
-	return images
+	return
 }

 // splitText splits text into chunks of at most maxLen bytes,