chore(server): improve transcript stability (#13821)

<!-- This is an auto-generated comment: release notes by coderabbit.ai
-->
## Summary by CodeRabbit

* **New Features**
* Enhanced audio/video detection for MP4 files to better distinguish
audio-only vs. video.

* **Dependencies**
* Added MP4 parsing dependency and updated AI provider libraries
(Anthropic, Google, OpenAI, etc.).

* **Bug Fixes**
  * Tightened authentication state validation for magic-link/OTP flows.
* Stricter space-join validation to reject invalid client
types/versions.
  * Improved transcript entry deduplication and data handling.

* **API**
* Transcript submit payload now requires infos and removes deprecated
url/mimeType fields.
<!-- end of auto-generated comment: release notes by coderabbit.ai -->

packages/backend/native/src/file_type.rs

@@ -1,12 +1,41 @@
+use mp4parse::{read_mp4, TrackType};
 use napi_derive::napi;
 
 #[napi]
 pub fn get_mime(input: &[u8]) -> String {
-  if let Some(kind) = infer::get(&input[..4096.min(input.len())]) {
+  let mimetype = if let Some(kind) = infer::get(&input[..4096.min(input.len())]) {
     kind.mime_type().to_string()
   } else {
     file_format::FileFormat::from_bytes(input)
       .media_type()
       .to_string()
+  };
+  if mimetype == "video/mp4" {
+    detect_mp4_flavor(input)
+  } else {
+    mimetype
+  }
+}
+
+fn detect_mp4_flavor(input: &[u8]) -> String {
+  let mut cursor = std::io::Cursor::new(input);
+  match read_mp4(&mut cursor) {
+    Ok(ctx) => {
+      let mut has_video = false;
+      let mut has_audio = false;
+      for track in ctx.tracks.iter() {
+        match track.track_type {
+          TrackType::Video | TrackType::AuxiliaryVideo | TrackType::Picture => has_video = true,
+          TrackType::Audio => has_audio = true,
+          _ => {}
+        }
+      }
+      if !has_video && has_audio {
+        "audio/m4a".to_string()
+      } else {
+        "video/mp4".to_string()
+      }
+    }
+    Err(_) => "video/mp4".to_string(),
   }
 }