fix(mobile): show the files a bot sends — video, audio, documents (MOCA-155)

android/app/src/main/kotlin/com/openmausbot/companion/ui/MessageRow.kt

@@ -89,6 +89,7 @@ import androidx.compose.ui.unit.sp
 import com.openmausbot.companion.audio.VoiceNoteController
 import com.openmausbot.companion.core.Chat
 import com.openmausbot.companion.core.AttachedMessageContent
+import com.openmausbot.companion.core.attachedFiles
 import com.openmausbot.companion.core.generatedImages
 import com.openmausbot.companion.core.voiceNotes
 import com.openmausbot.companion.core.DisplayedMessageAttachment
@@ -546,6 +547,14 @@ private fun TextBubble(
             message.generatedImages.forEach { attachment ->
                 SharedAttachmentView(threadId, message, attachment, openAttachment)
             }
+            // Documents, audio and video a bot sent with attach_file (MOCA-155).
+            // The card opens the file sheet; Open hands video to the player.
+            message.attachedFiles.forEach { attachment ->
+                SharedAttachmentView(
+                    threadId, message, attachment, openAttachment,
+                    foreground = if (mine) BubbleColor.mineText else MaterialTheme.colorScheme.onSurface,
+                )
+            }
             // Bots get markdown, you do not — the same split the desktop makes.
             // Markdown you did not intend is worse than markdown you did: a
             // message about `**` should show the asterisks.
@@ -597,16 +606,19 @@ private fun SharedAttachmentView(
     message: Message,
     attachment: DisplayedMessageAttachment,
     onOpen: ((DisplayedMessageAttachment, Message, DownloadedFile?) -> Unit)?,
+    /** Your own bubble is blue with white text; a bot's file sits on the theme surface. */
+    foreground: Color = BubbleColor.mineText,
 ) {
     if (attachment.kind == DisplayedMessageAttachment.Kind.IMAGE) {
         SharedImageAttachment(threadId, message, attachment, onOpen)
         return
     }
+    val family = attachment.fileFamily
     Row(
         modifier = Modifier
             .widthIn(max = 360.dp)
             .clip(RoundedCornerShape(16.dp))
-            .background(BubbleColor.mineText.copy(alpha = 0.10f))
+            .background(foreground.copy(alpha = 0.10f))
             .clickable(enabled = onOpen != null, role = Role.Button) {
                 onOpen?.invoke(attachment, message, null)
             }
@@ -617,17 +629,27 @@ private fun SharedAttachmentView(
         horizontalArrangement = Arrangement.spacedBy(10.dp),
         verticalAlignment = Alignment.CenterVertically,
     ) {
-        Text(stringResource(R.string.mobile_file_b4915d3a), fontSize = 11.sp, fontWeight = FontWeight.Bold, color = BubbleColor.mineText.copy(alpha = 0.68f))
+        val label = when (family) {
+            DisplayedMessageAttachment.FileFamily.VIDEO -> R.string.mobile_file_video
+            DisplayedMessageAttachment.FileFamily.AUDIO -> R.string.mobile_file_audio
+            DisplayedMessageAttachment.FileFamily.DOCUMENT -> R.string.mobile_file_b4915d3a
+        }
+        val hint = if (family == DisplayedMessageAttachment.FileFamily.DOCUMENT) {
+            R.string.mobile_tap_to_preview_fa5ce0ea
+        } else {
+            R.string.mobile_tap_to_play
+        }
+        Text(stringResource(label), fontSize = 11.sp, fontWeight = FontWeight.Bold, color = foreground.copy(alpha = 0.68f))
         Column(modifier = Modifier.weight(1f)) {
             Text(
                 attachment.name,
                 fontSize = 14.sp,
                 fontWeight = FontWeight.Medium,
-                color = BubbleColor.mineText,
+                color = foreground,
                 maxLines = 1,
                 overflow = TextOverflow.Ellipsis,
             )
-            Text(stringResource(R.string.mobile_tap_to_preview_fa5ce0ea), fontSize = 12.sp, color = BubbleColor.mineText.copy(alpha = 0.68f))
+            Text(stringResource(hint), fontSize = 12.sp, color = foreground.copy(alpha = 0.68f))
         }
     }
 }

android/app/src/main/res/values-b+zh+Hans/strings.xml

@@ -109,6 +109,9 @@
     <string name="mobile_expand_9869e506">展开</string>
     <string name="mobile_expanded_6d170474">已展开</string>
     <string name="mobile_file_b4915d3a">文件</string>
+    <string name="mobile_file_video">视频</string>
+    <string name="mobile_file_audio">音频</string>
+    <string name="mobile_tap_to_play">点一下以播放</string>
     <string name="mobile_for_example_summarize_this_and_lis_12b3d099">例如：整理摘要并列出后续步骤</string>
     <string name="mobile_from_5_minutes_to_24_hours_a6bafa40">从 5 分钟到 24 小时</string>
     <string name="mobile_group_name_optional_0fd3a15c">群组名称（选填）</string>

android/app/src/main/res/values-b+zh+Hant/strings.xml

@@ -109,6 +109,9 @@
     <string name="mobile_expand_9869e506">展開</string>
     <string name="mobile_expanded_6d170474">已展開</string>
     <string name="mobile_file_b4915d3a">檔案</string>
+    <string name="mobile_file_video">影片</string>
+    <string name="mobile_file_audio">音訊</string>
+    <string name="mobile_tap_to_play">點一下以播放</string>
     <string name="mobile_for_example_summarize_this_and_lis_12b3d099">例如：整理摘要並列出後續步驟</string>
     <string name="mobile_from_5_minutes_to_24_hours_a6bafa40">從 5 分鐘到 24 小時</string>
     <string name="mobile_group_name_optional_0fd3a15c">群組名稱（選填）</string>

android/app/src/main/res/values/strings.xml

@@ -110,6 +110,9 @@
     <string name="mobile_expand_9869e506">Expand</string>
     <string name="mobile_expanded_6d170474">Expanded</string>
     <string name="mobile_file_b4915d3a">FILE</string>
+    <string name="mobile_file_video">VIDEO</string>
+    <string name="mobile_file_audio">AUDIO</string>
+    <string name="mobile_tap_to_play">Tap to play</string>
     <string name="mobile_for_example_summarize_this_and_lis_12b3d099">For example: summarize this and list the next steps</string>
     <string name="mobile_from_5_minutes_to_24_hours_a6bafa40">From 5 minutes to 24 hours</string>
     <string name="mobile_group_name_optional_0fd3a15c">Group name (optional)</string>

android/core/src/main/kotlin/com/openmausbot/companion/core/AttachedMessageContent.kt

@@ -10,6 +10,20 @@ data class DisplayedMessageAttachment(
     val durationMs: Double? = null,
 ) {
     enum class Kind { IMAGE, FILE, AUDIO }
+
+    /**
+     * What a file card promises: video and audio play, the rest preview. Read
+     * from the extension, the signal the server uses to pick the file's type
+     * (mimeFor in server/message-file.ts). Port of the iOS `fileFamily`.
+     */
+    enum class FileFamily { VIDEO, AUDIO, DOCUMENT }
+
+    val fileFamily: FileFamily
+        get() = when (name.substringAfterLast('.', "").lowercase()) {
+            "mp4", "mov", "m4v", "webm" -> FileFamily.VIDEO
+            "mp3", "m4a", "wav", "aac", "ogg", "oga", "opus", "flac" -> FileFamily.AUDIO
+            else -> FileFamily.DOCUMENT
+        }
 }
 
 data class AttachedMessageContent(
@@ -212,6 +226,25 @@ data class AttachedMessageContent(
     }
 }
 
+/**
+ * Files a bot sent with attach_file — documents, audio and video — in wire
+ * order and deduplicated like [generatedImages]. Before these rendered, a
+ * file-only reply was an empty bubble on the phone (MOCA-155). The path goes
+ * unchanged to the originating message's authenticated file route.
+ */
+val Message.attachedFiles: List<DisplayedMessageAttachment>
+    get() = attachments.orEmpty()
+        .filter { it.kind == "file" && !it.path.isNullOrBlank() }
+        .distinctBy { it.path }
+        .map {
+            val path = requireNotNull(it.path)
+            DisplayedMessageAttachment(
+                kind = DisplayedMessageAttachment.Kind.FILE,
+                path = path,
+                name = AttachedMessageContent.displayName(it.name, path, DisplayedMessageAttachment.Kind.FILE),
+            )
+        }
+
 /** Preserve the server path exactly for the originating message's authenticated file route. */
 val Message.generatedImages: List<DisplayedMessageAttachment>
     get() = attachments.orEmpty()

android/core/src/main/kotlin/com/openmausbot/companion/core/ChatPreferences.kt

@@ -137,7 +137,10 @@ fun rosterPreview(messages: List<Message>, detail: ActivityDetail): String =
 
 /** What a single message reads as in a roster row. */
 internal fun previewText(message: Message): String = when (message.kind) {
-    Message.Kind.TEXT -> message.webhookContent?.task ?: message.text.orEmpty()
+    Message.Kind.TEXT -> message.webhookContent?.task
+        ?: message.text?.takeIf { it.isNotEmpty() }
+        // A bot that only sent a file says so by its name.
+        ?: message.attachedFiles.firstOrNull()?.name.orEmpty()
     // a pending card's question is the preview; the roster row already says
     // "waiting on you" beside it
     Message.Kind.OPTIONS -> {

android/core/src/main/kotlin/com/openmausbot/companion/core/Models.kt

@@ -1532,12 +1532,18 @@ data class ServerSession(val id: String, val label: String, val scopes: List<Str
 @Serializable
 data class ServerEnvironment(val environmentId: String, val label: String)
 
-/** Unknown attachment kinds remain decodable and are not rendered. */
+/**
+ * Image entries display inline, file entries (a bot's attach_file: documents,
+ * audio, video) show as file cards ([attachedFiles]), audio entries render as
+ * voice notes. Unknown attachment kinds remain decodable and are not rendered.
+ */
 @Serializable
 data class MessageImageAttachment(
     val kind: String,
     val path: String? = null,
     val mime: String? = null,
     /** `kind == "audio"`: the server's duration estimate, used until the player loads metadata. */
     val durationMs: Double? = null,
+    /** The file's name as the bot sent it, for `kind == "file"`. Presentation only; basenamed before display. */
+    val name: String? = null,
 )

android/core/src/test/kotlin/com/openmausbot/companion/core/GeneratedImageTest.kt

@@ -35,4 +35,47 @@ class GeneratedImageTest {
         assertEquals(emptyList(), message.generatedImages)
         assertEquals("Still visible", message.text)
     }
+
+    /**
+     * A bot's attach_file sends documents, audio and video as `kind:"file"`
+     * with a name (server/bot-attachment.ts). The phone dropped them, and the
+     * message read as an empty bubble (MOCA-155). Port of the iOS test.
+     */
+    @Test
+    fun aBotsFileAttachmentsBecomeNamedFileCards() {
+        val source = """{"id":"sent","role":"bot","kind":"text","at":1,"text":"","attachments":[{"kind":"file","path":"/data/attachments/9f.mp4","mime":"video/mp4","name":"demo clip.mp4"},{"kind":"image","path":"/data/attachments/a1.png","mime":"image/png"},{"kind":"file","path":"/data/attachments/77.pdf","mime":"application/pdf","name":"../../report.pdf"},{"kind":"file","path":"/data/attachments/9f.mp4","name":"again.mp4"},{"kind":"file","path":" "},{"kind":"hologram","path":"/data/attachments/x.bin"}]}"""
+        val message = CompanionJson.decodeFromString<Message>(source)
+        assertEquals(
+            listOf(
+                DisplayedMessageAttachment(DisplayedMessageAttachment.Kind.FILE, "demo clip.mp4", "/data/attachments/9f.mp4"),
+                // A crafted name is presentation only and never a path.
+                DisplayedMessageAttachment(DisplayedMessageAttachment.Kind.FILE, "report.pdf", "/data/attachments/77.pdf"),
+            ),
+            message.attachedFiles,
+        )
+        assertEquals(listOf("/data/attachments/a1.png"), message.generatedImages.map { it.path }, "images keep their own card")
+        val cached = CompanionJson.decodeFromString<Message>(CompanionJson.encodeToString(message))
+        assertEquals(message.attachedFiles, cached.attachedFiles)
+    }
+
+    @Test
+    fun aFileOnlyBotMessagePreviewsAsItsFileName() {
+        val message = CompanionJson.decodeFromString<Message>(
+            """{"id":"sent","role":"bot","kind":"text","at":1,"text":"","attachments":[{"kind":"file","path":"/data/attachments/9f.mp4","mime":"video/mp4","name":"demo clip.mp4"}]}""",
+        )
+        assertEquals("demo clip.mp4", rosterPreview(listOf(message), ActivityDetail.FULL))
+        assertEquals("Here is the clip", rosterPreview(listOf(message.copy(text = "Here is the clip")), ActivityDetail.FULL))
+    }
+
+    @Test
+    fun aFileCardKnowsVideoAndAudioFromTheExtension() {
+        fun family(name: String) = DisplayedMessageAttachment(DisplayedMessageAttachment.Kind.FILE, name, "/data/attachments/x").fileFamily
+        assertEquals(DisplayedMessageAttachment.FileFamily.VIDEO, family("demo clip.MP4"))
+        assertEquals(DisplayedMessageAttachment.FileFamily.VIDEO, family("screen.mov"))
+        assertEquals(DisplayedMessageAttachment.FileFamily.VIDEO, family("talk.webm"))
+        assertEquals(DisplayedMessageAttachment.FileFamily.AUDIO, family("memo.m4a"))
+        assertEquals(DisplayedMessageAttachment.FileFamily.AUDIO, family("song.mp3"))
+        assertEquals(DisplayedMessageAttachment.FileFamily.DOCUMENT, family("weekly-report.pdf"))
+        assertEquals(DisplayedMessageAttachment.FileFamily.DOCUMENT, family("notes"))
+    }
 }

ios/App/AttachmentViews.swift

@@ -234,10 +234,18 @@ struct TranscriptAttachmentView: View {
         .accessibilityHint(thumbnail == nil ? "Loads the image preview" : "Opens the image full screen")
     }
 
+    private var fileSymbol: String {
+        switch attachment.fileFamily {
+        case .video: "play.rectangle.fill"
+        case .audio: "waveform"
+        case .document: "doc.fill"
+        }
+    }
+
     private var fileCard: some View {
         Button(action: openPreview) {
             HStack(spacing: 10) {
-                Image(systemName: "doc.fill")
+                Image(systemName: fileSymbol)
                     .font(.system(size: 16, weight: .medium))
                     .foregroundStyle(foreground)
                     .frame(width: 38, height: 38)
@@ -248,7 +256,7 @@ struct TranscriptAttachmentView: View {
                     Text(attachment.name)
                         .font(.system(size: 13, weight: .semibold))
                         .lineLimit(1)
-                    Text(previewLoading ? "Opening…" : "Tap to preview")
+                    Text(previewLoading ? "Opening…" : attachment.fileFamily == .document ? "Tap to preview" : "Tap to play")
                         .font(.system(size: 11))
                         .foregroundStyle(foreground.opacity(0.68))
                 }

ios/App/ChatView.swift

@@ -1883,6 +1883,14 @@ struct TextBubble: View {
                         messageId: message.id, foreground: mine ? BubbleColor.mineText : .primary
                     )
                 }
+                // Documents, audio and video a bot sent with attach_file. The
+                // card opens the full-screen viewer, which plays video.
+                ForEach(message.attachedFiles, id: \.path) { attachment in
+                    TranscriptAttachmentView(
+                        attachment: attachment, threadId: chat.threadId,
+                        messageId: message.id, foreground: mine ? BubbleColor.mineText : .primary
+                    )
+                }
                 // Bots get markdown, you do not — the same split the desktop
                 // makes. Markdown you did not intend is worse than markdown
                 // you did: a message about `**` should show the asterisks.

ios/App/Localizable.xcstrings

@@ -11197,6 +11197,16 @@
         }
       }
     },
+    "Tap to play": {
+      "localizations": {
+        "pt-BR": {
+          "stringUnit": {
+            "state": "translated",
+            "value": "Toque para reproduzir"
+          }
+        }
+      }
+    },
     "Nothing scheduled": {
       "localizations": {
         "pt-BR": {

ios/Sources/CompanionCore/AttachedMessageContent.swift

@@ -21,6 +21,22 @@ public struct DisplayedMessageAttachment: Hashable, Sendable {
         self.path = path
         self.name = name
     }
+
+    /// What a file card promises: video and audio play in the viewer, the
+    /// rest preview. Read from the extension, the same signal the server
+    /// uses to pick the file's type (mimeFor in server/message-file.ts).
+    public enum FileFamily: Sendable {
+        case video, audio, document
+    }
+
+    public var fileFamily: FileFamily {
+        let ext = (name as NSString).pathExtension.lowercased()
+        switch ext {
+        case "mp4", "mov", "m4v", "webm": return .video
+        case "mp3", "m4a", "wav", "aac", "ogg", "oga", "opus", "flac": return .audio
+        default: return .document
+        }
+    }
 }
 
 public struct AttachedMessageContent: Hashable, Sendable {
@@ -294,6 +310,23 @@ public struct AttachedMessageContent: Hashable, Sendable {
 }
 
 extension Message {
+    /// Files a bot sent with attach_file — documents, audio and video — in wire
+    /// order and deduplicated like generatedImages. Before these rendered, a
+    /// file-only reply was an empty bubble on the phone (MOCA-155). Paths are
+    /// server metadata, passed unchanged to the message-scoped file route.
+    public var attachedFiles: [DisplayedMessageAttachment] {
+        var seen = Set<String>()
+        return (attachments ?? []).compactMap { attachment in
+            guard attachment.kind == "file", let path = attachment.path,
+                  !path.trimmingCharacters(in: .whitespacesAndNewlines).isEmpty,
+                  seen.insert(path).inserted else { return nil }
+            return DisplayedMessageAttachment(
+                kind: .file, path: path,
+                name: AttachedMessageContent.displayName(providedName: attachment.name, path: path, kind: .file)
+            )
+        }
+    }
+
     /// Paths are server metadata, passed unchanged to the message-scoped file route.
     public var generatedImages: [DisplayedMessageAttachment] {
         var seen = Set<String>()

ios/Sources/CompanionCore/ChatPreferences.swift

@@ -180,7 +180,11 @@ public func rosterPreview(_ messages: [Message], detail: ActivityDetail) -> Stri
 /// What a single message reads as in a roster row.
 func previewText(of message: Message) -> String {
     switch message.kind {
-    case .text: return message.webhookContent?.task ?? message.text ?? ""
+    case .text:
+        if let task = message.webhookContent?.task { return task }
+        if let text = message.text, !text.isEmpty { return text }
+        // A bot that only sent a file says so by its name.
+        return message.attachedFiles.first?.name ?? ""
     // a pending card's question is the preview; the roster row already
     // says "waiting on you" beside it
     case .options:

ios/Sources/CompanionCore/Models.swift

@@ -1708,13 +1708,19 @@ public struct ServerEnvironment: Codable, Hashable, Sendable {
     public var version: String?
 }
 
-/// Keep future attachment kinds decodable; image entries display inline and
-/// audio entries render as voice notes (Message.voiceNotes). Unknown kinds
-/// decode without breaking, so a newer computer never gaps the transcript.
+/// Keep future attachment kinds decodable; image entries display inline,
+/// file entries (a bot's attach_file: documents, audio, video) show as file
+/// cards (Message.attachedFiles), and audio entries render as voice notes
+/// (Message.voiceNotes). Unknown kinds decode without breaking, so a newer
+/// computer never gaps the transcript.
 public struct MessageImageAttachment: Codable, Hashable, Sendable {
     public var kind: String
     public var path: String?
     public var mime: String?
+    /// The file's name as the bot sent it, for a `kind: "file"` entry.
+    /// Presentation only: it is basenamed before display and never used as
+    /// a path.
+    public var name: String?
     /// The server's duration estimate for an audio attachment, in
     /// milliseconds; shown until the player loads real metadata.
     public var durationMs: Double?