feat: Enhance video composition functionality with detailed docstring and improved command handling

app.py

@@ -190,19 +190,29 @@ YOUR RESPONSE:
             # Find content between ```sh or ```bash and the next ```
             import re
 
-            command_match = re.search(r"```(?:sh|bash)?\n(.*?)\n```", content, re.DOTALL)
+            command_match = re.search(
+                r"```(?:sh|bash)?\n(.*?)\n```", content, re.DOTALL
+            )
             if command_match:
                 command = command_match.group(1).strip()
             else:
                 # Try to find a line that starts with ffmpeg
-                ffmpeg_lines = [line.strip() for line in content.split('\n') if line.strip().startswith('ffmpeg')]
+                ffmpeg_lines = [
+                    line.strip()
+                    for line in content.split("\n")
+                    if line.strip().startswith("ffmpeg")
+                ]
                 if ffmpeg_lines:
                     command = ffmpeg_lines[0]
                 else:
                     command = content.replace("\n", "")
         else:
             # Try to find a line that starts with ffmpeg
-            ffmpeg_lines = [line.strip() for line in content.split('\n') if line.strip().startswith('ffmpeg')]
+            ffmpeg_lines = [
+                line.strip()
+                for line in content.split("\n")
+                if line.strip().startswith("ffmpeg")
+            ]
             if ffmpeg_lines:
                 command = ffmpeg_lines[0]
             else:
@@ -220,13 +230,40 @@ YOUR RESPONSE:
 def execute_ffmpeg_command(args, temp_dir, output_file_path):
     """Execute FFmpeg command with GPU acceleration"""
     final_command = args + ["-y", output_file_path]
-    print(
-        f"\n=== EXECUTING FFMPEG COMMAND ===\nffmpeg {' '.join(final_command[1:])}\n"
-    )
+    print(f"\n=== EXECUTING FFMPEG COMMAND ===\nffmpeg {' '.join(final_command[1:])}\n")
     subprocess.run(final_command, cwd=temp_dir)
     return output_file_path
 
 
+def compose_video(
+    prompt: str,
+    files: list = None,
+    top_p: float = 0.7,
+    temperature: float = 0.1,
+    model_choice: str = "deepseek-ai/DeepSeek-V3",
+) -> str:
+    """
+    Compose a video from media assets using natural language instructions.
+
+    This tool generates FFmpeg commands using AI and executes them to create videos
+    from uploaded images, videos, and audio files based on natural language descriptions.
+
+    Args:
+        prompt (str): Natural language instructions for video composition (e.g., "Create a slideshow with background music")
+        files (list, optional): List of media files (images, videos, audio) to use
+        top_p (float): Top-p sampling parameter for AI model (0.0-1.0, default: 0.7)
+        temperature (float): Temperature parameter for AI model creativity (0.0-5.0, default: 0.1)
+        model_choice (str): AI model to use for command generation (default: "deepseek-ai/DeepSeek-V3")
+
+    Returns:
+        str: Path to the generated video file
+
+    Example:
+        compose_video("Create a 10-second slideshow from the images with fade transitions", files=[img1, img2, img3])
+    """
+    return update(files or [], prompt, top_p, temperature, model_choice)
+
+
 def update(
     files,
     prompt,
@@ -291,7 +328,9 @@ def update(
             execute_ffmpeg_command(args, temp_dir, output_file_path)
             # Extract just the command for display
             command_for_display = f"ffmpeg {' '.join(args[1:])} -y output.mp4"
-            generated_command = f"### Generated Command\n```bash\n{command_for_display}\n```"
+            generated_command = (
+                f"### Generated Command\n```bash\n{command_for_display}\n```"
+            )
             return output_file_path, gr.update(value=generated_command)
         except Exception as e:
             attempts += 1
@@ -300,6 +339,26 @@ def update(
                 raise gr.Error(e)
 
 
+# Create MCP-compatible interface
+mcp_interface = gr.Interface(
+    fn=compose_video,
+    inputs=[
+        gr.Textbox(
+            value="Create a slideshow with background music",
+            label="Video Composition Instructions",
+        ),
+        gr.File(file_count="multiple", label="Media Files", file_types=allowed_medias),
+        gr.Slider(0.0, 1.0, value=0.7, label="Top-p"),
+        gr.Slider(0.0, 5.0, value=0.1, label="Temperature"),
+        gr.Radio(
+            choices=list(MODELS.keys()), value=list(MODELS.keys())[0], label="Model"
+        ),
+    ],
+    outputs=gr.Video(label="Generated Video"),
+    title="AI Video Composer MCP Tool",
+    description="Compose videos from media assets using natural language",
+)
+
 with gr.Blocks() as demo:
     gr.Markdown(
         """
@@ -358,13 +417,13 @@ with gr.Blocks() as demo:
             examples=[
                 [
                     ["./examples/ai_talk.wav", "./examples/bg-image.png"],
-                   "Use the image as the background with a waveform visualization for the audio positioned in center of the video.",
-                   0.7,
-                   0.1,
-                   list(MODELS.keys())[0],
-               ],
-               [
-                   ["./examples/ai_talk.wav", "./examples/bg-image.png"],
+                    "Use the image as the background with a waveform visualization for the audio positioned in center of the video.",
+                    0.7,
+                    0.1,
+                    list(MODELS.keys())[0],
+                ],
+                [
+                    ["./examples/ai_talk.wav", "./examples/bg-image.png"],
                     "Use the image as the background with a waveform visualization for the audio positioned in center of the video. Make sure the waveform has a max height of 250 pixels.",
                     0.7,
                     0.1,
@@ -380,13 +439,13 @@ with gr.Blocks() as demo:
                         "./examples/cat6.jpeg",
                         "./examples/heat-wave.mp3",
                     ],
-                   "Create a 3x2 grid of the cat images with the audio as background music. Make the video duration match the audio duration.",
-                   0.7,
-                   0.1,
-                   list(MODELS.keys())[0],
-               ],
-           ],
-           inputs=[user_files, user_prompt, top_p, temperature, model_choice],
+                    "Create a 3x2 grid of the cat images with the audio as background music. Make the video duration match the audio duration.",
+                    0.7,
+                    0.1,
+                    list(MODELS.keys())[0],
+                ],
+            ],
+            inputs=[user_files, user_prompt, top_p, temperature, model_choice],
             outputs=[generated_video, generated_command],
             fn=update,
             run_on_click=True,
@@ -402,5 +461,9 @@ with gr.Blocks() as demo:
             """,
         )
 
+# Launch MCP interface for tool access
+mcp_interface.queue(default_concurrency_limit=200)
+
+# Launch main demo
 demo.queue(default_concurrency_limit=200)
-demo.launch(show_api=False, ssr_mode=False)
+demo.launch(show_api=False, ssr_mode=False, mcp_server=True)