feat: add auto-select from preferences and type picker
All checks were successful
CI / build (push) Successful in 52s

- Add MediaTypeAsk (0) as the default, mapping old video_audio to Ask
- Show Video/Audio/Ask type picker when no default media type is set
- Auto-select matching format and skip picker when defaults are configured
- Add autoSelectVideoFormat / autoSelectAudioFormat helpers with closest-match
- Add buildFilteredFormatList for type-filtered format lists
- Add Media Type setting in settings UI (Ask/Video/Audio)
- Add --merge-output-format mp4 for video downloads via yt-dlp
- Fix formatLabelForDisplay to show Height fallback and Extension
- Add cookies volume mount to docker-compose.yml
- Document cookies setup and new features in README
This commit is contained in:
2026-06-28 11:28:53 +03:30
parent 6bbc87a7c8
commit b11b5649a5
12 changed files with 578 additions and 18 deletions

View File

@@ -24,6 +24,7 @@ type stepState struct {
URL string
MediaInfo *domain.MediaInfo
Step int
SelectedType string // "video", "audio", "ask", or "" when not yet chosen
MainFormatID string
MainFormat *domain.Format
SecondaryID string
@@ -158,6 +159,34 @@ func (h *Handler) handleURLInput(ctx context.Context, chatID int64, text string,
h.sendMediaPreview(ctx, chatID, info)
prefs, prefErr := h.Repo.GetOrCreatePreferences(userID)
if prefErr == nil && prefs.DefaultMediaType != domain.MediaTypeAsk {
h.autoSelectAndDownload(ctx, chatID, userID, url, info, prefs)
return
}
ss := &stepState{
URL: url,
MediaInfo: info,
Step: 0,
CreatedAt: time.Now(),
}
h.setUserStepState(chatID, ss)
kb := mediaTypeKeyboard()
h.sendWithKB(ctx, chatID, "Select media type:", kb)
}
// autoSelectAndDownload picks the best matching format from preferences and
// skips the picker entirely.
func (h *Handler) autoSelectAndDownload(ctx context.Context, chatID int64, userID int64, url string, info *domain.MediaInfo, prefs *domain.UserPreferences) {
msg, err := h.Bot.SendMessage(ctx, &tgbot.SendMessageParams{ChatID: chatID, Text: "Processing..."})
if err != nil {
slog.Warn("send processing message failed", "error", err)
return
}
msgID := msg.ID
ss := &stepState{
URL: url,
MediaInfo: info,
@@ -165,14 +194,74 @@ func (h *Handler) handleURLInput(ctx context.Context, chatID int64, text string,
CreatedAt: time.Now(),
}
ss.FormatIDs, _ = buildFormatList(info.Formats)
var mainID, secondaryID string
switch prefs.DefaultMediaType {
case domain.MediaTypeVideo:
mainID, secondaryID = autoSelectVideoFormat(info.Formats, prefs.DefaultVideoFormat)
case domain.MediaTypeAudio:
mainID = autoSelectAudioFormat(info.Formats, prefs.DefaultAudioFormat)
default:
mainID = "best"
}
if mainID == "" {
mainID = "best"
}
ss.MainFormatID = mainID
if mainID == "best" {
if info.IsAudioOnlyContent() {
ss.MainFormat = &domain.Format{ID: "best", HasAudio: true, IsAudioOnly: true}
} else {
ss.MainFormat = &domain.Format{ID: "best", HasVideo: true, HasAudio: true}
}
} else {
for i := range info.Formats {
f := &info.Formats[i]
if f.ID == mainID {
ss.MainFormat = f
break
}
}
}
if secondaryID != "" {
ss.SecondaryID = secondaryID
for i := range info.Formats {
f := &info.Formats[i]
if f.ID == secondaryID {
ss.SecondaryFmt = f
break
}
}
}
h.advanceToLanguage(ctx, chatID, msgID, userID, ss)
}
// handleTypeSelection processes the initial type picker (Video/Audio/Ask).
func (h *Handler) handleTypeSelection(ctx context.Context, chatID int64, msgID int, userID int64, mediaType string) {
ss := h.getUserStepState(chatID)
if ss == nil {
slog.Warn("nil stepState in handleTypeSelection", "chat_id", chatID, "user_id", userID)
return
}
ss.SelectedType = mediaType
ids, err := buildFilteredFormatList(ss.MediaInfo.Formats, mediaType)
if err != nil {
h.editText(ctx, chatID, msgID, "No formats available for this type.", nil)
return
}
ss.FormatIDs = ids
h.setUserStepState(chatID, ss)
labels := make([]string, len(ss.FormatIDs))
for i, id := range ss.FormatIDs {
labels := make([]string, len(ids))
for i, id := range ids {
if id == "best" {
labels[i] = "Best"
} else {
f := findFormatByID(info.Formats, id)
f := findFormatByID(ss.MediaInfo.Formats, id)
if f != nil {
labels[i] = formatLabelForDisplay(*f)
} else {
@@ -180,8 +269,8 @@ func (h *Handler) handleURLInput(ctx context.Context, chatID int64, text string,
}
}
}
kb := formatKeyboard("format_", labels, ss.FormatIDs)
h.sendWithKB(ctx, chatID, "Select format:", kb)
kb := formatKeyboard("format_", labels, ids)
h.editText(ctx, chatID, msgID, "Select format:", &kb)
}
// handleFormatSelection processes a format pick from the unified list.
@@ -612,13 +701,22 @@ func (h *Handler) handleBackStep(ctx context.Context, chatID int64, msgID int, c
return
}
// If on the format list with a selected type, go back to type picker.
if ss.Step == 0 && ss.SelectedType != "" {
ss.SelectedType = ""
h.setUserStepState(chatID, ss)
kb := mediaTypeKeyboard()
h.editText(ctx, chatID, msgID, "Select media type:", &kb)
return
}
ss.Step--
switch ss.Step {
case -1:
h.clearStepState(chatID)
h.handleDownloadPrompt(ctx, chatID, msgID, cbID, userID)
case 0:
ss.FormatIDs, _ = buildFormatList(ss.MediaInfo.Formats)
ss.FormatIDs, _ = buildFilteredFormatList(ss.MediaInfo.Formats, "")
labels := make([]string, len(ss.FormatIDs))
for i, id := range ss.FormatIDs {
if id == "best" {
@@ -685,6 +783,162 @@ func buildFormatList(formats []domain.Format) ([]string, error) {
return ids, nil
}
// buildFilteredFormatList returns format IDs filtered by media type for the type picker.
// When mediaType is "video", only formats with video are returned.
// When mediaType is "audio", only audio-only formats are returned.
// When empty or "ask", all formats are returned (current behavior).
func buildFilteredFormatList(formats []domain.Format, mediaType string) ([]string, error) {
deduped := dl.DeduplicateResolutions(formats)
ids := make([]string, 0, len(deduped)+1)
ids = append(ids, "best")
switch mediaType {
case "video":
for _, f := range deduped {
if f.HasVideo {
ids = append(ids, f.ID)
}
}
case "audio":
for _, f := range deduped {
if f.HasAudio && f.IsAudioOnly {
ids = append(ids, f.ID)
}
}
default:
for _, f := range deduped {
if f.HasVideo {
ids = append(ids, f.ID)
}
}
for _, f := range deduped {
if !f.HasVideo {
ids = append(ids, f.ID)
}
}
}
return ids, nil
}
// autoSelectVideoFormat finds the closest video format to the preferred resolution.
// Returns the main format ID and optionally a secondary audio ID if the best match
// is video-only. Prefers formats at or above the target resolution, falling back to
// the closest below. Returns ("best", "") if the preference is "best" or no match.
func autoSelectVideoFormat(formats []domain.Format, preferredID string) (mainID, secondaryID string) {
if preferredID == "" || preferredID == "best" {
return "best", ""
}
var targetHeight int
if n, err := fmt.Sscanf(preferredID, "%dp", &targetHeight); err != nil || n != 1 {
return "best", ""
}
var bestFmt *domain.Format
bestDiff := math.MaxInt32
for i := range formats {
f := &formats[i]
if !f.HasVideo || f.Height == 0 {
continue
}
diff := f.Height - targetHeight
if diff >= 0 && diff < bestDiff {
bestDiff = diff
bestFmt = f
}
}
if bestFmt == nil {
bestDiff = math.MaxInt32
for i := range formats {
f := &formats[i]
if !f.HasVideo || f.Height == 0 {
continue
}
diff := targetHeight - f.Height
if diff >= 0 && diff < bestDiff {
bestDiff = diff
bestFmt = f
}
}
}
if bestFmt == nil {
return "best", ""
}
mainID = bestFmt.ID
if bestFmt.HasVideo && !bestFmt.HasAudio {
for i := range formats {
f := &formats[i]
if f.HasAudio && f.IsAudioOnly {
secondaryID = f.ID
break
}
}
}
return
}
// autoSelectAudioFormat finds the closest audio format to the preferred bitrate.
// Returns the format ID, or "best" if the preference is "best" or no match.
func autoSelectAudioFormat(formats []domain.Format, preferredID string) string {
if preferredID == "" || preferredID == "best" {
return "best"
}
var targetBitrate int
if n, err := fmt.Sscanf(preferredID, "%dk", &targetBitrate); err != nil || n != 1 {
return "best"
}
var bestFmt *domain.Format
bestDiff := math.MaxInt32
for i := range formats {
f := &formats[i]
if !f.HasAudio || !f.IsAudioOnly {
continue
}
var bitrate int
if n, err := fmt.Sscanf(f.Bitrate, "%dk", &bitrate); err != nil || n != 1 {
continue
}
diff := bitrate - targetBitrate
if diff >= 0 && diff < bestDiff {
bestDiff = diff
bestFmt = f
}
}
if bestFmt == nil {
bestDiff = math.MaxInt32
for i := range formats {
f := &formats[i]
if !f.HasAudio || !f.IsAudioOnly {
continue
}
var bitrate int
if n, err := fmt.Sscanf(f.Bitrate, "%dk", &bitrate); err != nil || n != 1 {
continue
}
diff := targetBitrate - bitrate
if diff >= 0 && diff < bestDiff {
bestDiff = diff
bestFmt = f
}
}
}
if bestFmt == nil {
return "best"
}
return bestFmt.ID
}
// findFormatByID returns the format with the given ID from the list, or nil.
func findFormatByID(formats []domain.Format, id string) *domain.Format {
for i := range formats {
@@ -716,6 +970,13 @@ func formatLabelForDisplay(f domain.Format) string {
}
return label
}
if f.Height > 0 {
label := fmt.Sprintf("%dp", f.Height)
if f.Filesize > 0 {
label += fmt.Sprintf(" (%s)", formatBytes(f.Filesize))
}
return label
}
if f.Bitrate != "" {
label := f.Bitrate
if f.Filesize > 0 {
@@ -723,7 +984,11 @@ func formatLabelForDisplay(f domain.Format) string {
}
return label
}
return f.ID
label := f.ID
if f.Extension != "" {
label += " (" + f.Extension + ")"
}
return label
}
// sendMediaPreview sends a thumbnail and metadata info before the selection flow.