Add API-friendly endpoint bypassing FileData validation

app.py

@@ -200,6 +200,64 @@ else:
     process_video_ui = _process_video_impl
 
 
+def _process_video_api_impl(
+    file_path, gazing_ratio, task_loss_requirement, output_fps, progress=None
+):
+    """API-friendly endpoint that takes a file path string instead of gr.File.
+    Returns only the gazing video path for programmatic use."""
+    if not file_path or not os.path.exists(file_path):
+        return "Error: file not found"
+
+    metadata = extract_metadata(file_path)
+    if metadata[1] is None:
+        return "Error: could not read file"
+
+    _, tmp_path, total_frames, fps, _, _ = metadata
+
+    yield "Loading model..."
+
+    if progress:
+        progress(0.0, desc="Loading model...")
+    setup = get_model(device)
+
+    yield "Processing video..."
+
+    if progress:
+        progress(0.1, desc="Processing video...")
+
+    def update_progress(pct, msg):
+        if progress:
+            progress(pct, desc=msg)
+
+    model_gazing_ratio = gazing_ratio * (196 / 265)
+
+    for results in process_video(
+        tmp_path,
+        setup,
+        gazing_ratio=model_gazing_ratio,
+        task_loss_requirement=task_loss_requirement,
+        progress_callback=update_progress,
+        spatial_batch_size=2,
+    ):
+        yield "Processing..."
+
+    yield "Saving output..."
+
+    fps_to_use = output_fps if output_fps is not None else results["fps"]
+
+    gazing_file = tempfile.NamedTemporaryFile(delete=False, suffix=".mp4")
+    save_video(results["gazing_frames"], gazing_file.name, fps_to_use)
+    gazing_file.close()
+
+    yield gazing_file.name
+
+
+if ZEROGPU_AVAILABLE:
+    process_video_api = spaces.GPU(duration=120)(_process_video_api_impl)
+else:
+    process_video_api = _process_video_api_impl
+
+
 def extract_first_frame_thumbnail(
     video_path, output_path, size=(200, 200), force=False
 ):
@@ -384,6 +442,33 @@ with gr.Blocks(title="AutoGaze Demo", delete_cache=(86400, 86400)) as demo:
         ],
     ).then(fn=cleanup_gpu, inputs=None, outputs=None)
 
+    # --- API-friendly endpoint (hidden tab, bypasses FileData validation) ---
+    with gr.Tab("API", visible=False):
+        api_file_path = gr.Textbox(label="File Path")
+        api_gazing_ratio = gr.Slider(
+            minimum=round(1 / 196, 2),
+            maximum=round(265 / 196, 2),
+            step=0.01,
+            value=0.75,
+            label="Gazing Ratio",
+        )
+        api_task_loss = gr.Slider(
+            minimum=0.0,
+            maximum=1.5,
+            step=0.05,
+            value=0.7,
+            label="Task Loss Requirement",
+        )
+        api_output_fps = gr.Number(label="Output FPS", value=None)
+        api_button = gr.Button("Process (API)")
+        api_result = gr.Textbox(label="Result Path")
+        api_button.click(
+            fn=process_video_api,
+            inputs=[api_file_path, api_gazing_ratio, api_task_loss, api_output_fps],
+            outputs=[api_result],
+            api_name="process_video_api",
+        )
+
     # Clean up GPU memory when user disconnects
     demo.unload(cleanup_gpu)