Fix Gradio 6 video input handling

app.py

@@ -49,7 +49,6 @@ def build_video_input(label: str) -> gr.Video:
     return gr.Video(
         label=label,
         sources=["upload", "webcam"],
-        type="filepath",
         format="mp4",
     )
 

signspeak/pipeline.py

@@ -28,15 +28,21 @@ def json_text(data: dict[str, Any]) -> str:
     return json.dumps(data, ensure_ascii=False, indent=2)
 
 
-def run_asl_video(video_file: str | None) -> tuple[str, dict[str, Any], str]:
+def run_asl_video(video_file: Any | None) -> tuple[str, dict[str, Any], str]:
     video_path = resolve_video_path(video_file)
     result = process_asl_video(video_path)
     intent = result["intent_input"]
     return json_text(intent), result, summarize_asl_result(result)
 
 
-def resolve_video_path(video_file: str | None) -> Path:
+def resolve_video_path(video_file: Any | None) -> Path:
     if video_file:
+        if isinstance(video_file, dict):
+            video_file = video_file.get("path") or video_file.get("name") or video_file.get("video")
+        elif hasattr(video_file, "path"):
+            video_file = video_file.path
+        elif hasattr(video_file, "name"):
+            video_file = video_file.name
         return Path(video_file)
     if DEFAULT_VIDEO_PATH.exists():
         return DEFAULT_VIDEO_PATH

tests/test_asl_pipeline.py

@@ -1,5 +1,5 @@
 from signspeak.asl.pipeline import build_intent_input
-from signspeak.pipeline import summarize_asl_result
+from signspeak.pipeline import resolve_video_path, summarize_asl_result
 
 
 def test_build_intent_input_matches_llm_schema():
@@ -36,3 +36,11 @@ def test_summarize_asl_result_is_stable_for_missing_fields():
     assert "ASL status: model_missing" in summary
     assert "Emotion: unknown (0.00)" in summary
 
+
+def test_resolve_video_path_accepts_gradio_dict_payload(tmp_path):
+    video_path = tmp_path / "capture.mp4"
+    video_path.write_bytes(b"demo")
+
+    resolved = resolve_video_path({"path": str(video_path)})
+
+    assert resolved == video_path