BladeSzaSza's picture
Upload folder using huggingface_hub
e49993e verified
"""
MCP (Model Context Protocol) Server for Laban Movement Analysis
Provides tools for video movement analysis accessible to AI agents
"""
import asyncio
import json
import os
import tempfile
from datetime import datetime
from pathlib import Path
from typing import Any, Dict, List, Optional, Tuple
from urllib.parse import urlparse
import aiofiles
import httpx
from mcp.server import Server
from mcp.server.stdio import stdio_server
from mcp.types import (
Tool,
TextContent,
ImageContent,
EmbeddedResource,
ToolParameterType,
ToolResponse,
ToolResult,
ToolError
)
# Add parent directory to path for imports
import sys
sys.path.insert(0, str(Path(__file__).parent))
from gradio_labanmovementanalysis import LabanMovementAnalysis
class LabanMCPServer:
"""MCP Server for Laban Movement Analysis"""
def __init__(self):
self.server = Server("laban-movement-analysis")
self.analyzer = LabanMovementAnalysis()
self.analysis_cache = {}
self.temp_dir = tempfile.mkdtemp(prefix="laban_mcp_")
# Register tools
self._register_tools()
def _register_tools(self):
"""Register all available tools"""
@self.server.tool()
async def analyze_video(
video_path: str,
model: str = "mediapipe",
enable_visualization: bool = False,
include_keypoints: bool = False
) -> ToolResult:
"""
Analyze movement in a video file using Laban Movement Analysis.
Args:
video_path: Path or URL to video file
model: Pose estimation model ('mediapipe', 'movenet', 'yolo')
enable_visualization: Generate annotated video output
include_keypoints: Include raw keypoint data in JSON
Returns:
Movement analysis results and optional visualization
"""
try:
# Handle URL vs local path
if video_path.startswith(('http://', 'https://')):
video_path = await self._download_video(video_path)
# Process video
json_output, viz_video = await asyncio.to_thread(
self.analyzer.process_video,
video_path,
model=model,
enable_visualization=enable_visualization,
include_keypoints=include_keypoints
)
# Store in cache
analysis_id = f"{Path(video_path).stem}_{datetime.now().isoformat()}"
self.analysis_cache[analysis_id] = {
"json_output": json_output,
"viz_video": viz_video,
"timestamp": datetime.now().isoformat()
}
# Format response
response_data = {
"analysis_id": analysis_id,
"analysis": json_output,
"visualization_path": viz_video if viz_video else None
}
return ToolResult(
success=True,
content=[TextContent(text=json.dumps(response_data, indent=2))]
)
except Exception as e:
return ToolResult(
success=False,
error=ToolError(message=f"Analysis failed: {str(e)}")
)
@self.server.tool()
async def get_analysis_summary(
analysis_id: str
) -> ToolResult:
"""
Get a human-readable summary of a previous analysis.
Args:
analysis_id: ID of the analysis to summarize
Returns:
Summary of movement analysis
"""
try:
if analysis_id not in self.analysis_cache:
return ToolResult(
success=False,
error=ToolError(message=f"Analysis ID '{analysis_id}' not found")
)
analysis_data = self.analysis_cache[analysis_id]["json_output"]
# Extract key information
summary = self._generate_summary(analysis_data)
return ToolResult(
success=True,
content=[TextContent(text=summary)]
)
except Exception as e:
return ToolResult(
success=False,
error=ToolError(message=f"Summary generation failed: {str(e)}")
)
@self.server.tool()
async def list_available_models() -> ToolResult:
"""
List available pose estimation models with their characteristics.
Returns:
Information about available models
"""
models_info = {
"mediapipe": {
"name": "MediaPipe Pose",
"keypoints": 33,
"dimensions": "3D",
"optimization": "CPU",
"best_for": "Single person, detailed analysis",
"speed": "Fast"
},
"movenet": {
"name": "MoveNet",
"keypoints": 17,
"dimensions": "2D",
"optimization": "Mobile/Edge",
"best_for": "Real-time applications, mobile devices",
"speed": "Very Fast"
},
"yolo": {
"name": "YOLO Pose",
"keypoints": 17,
"dimensions": "2D",
"optimization": "GPU",
"best_for": "Multi-person detection",
"speed": "Fast (with GPU)"
}
}
return ToolResult(
success=True,
content=[TextContent(text=json.dumps(models_info, indent=2))]
)
@self.server.tool()
async def batch_analyze(
video_paths: List[str],
model: str = "mediapipe",
parallel: bool = True
) -> ToolResult:
"""
Analyze multiple videos in batch.
Args:
video_paths: List of video paths or URLs
model: Pose estimation model to use
parallel: Process videos in parallel
Returns:
Batch analysis results
"""
try:
results = {}
if parallel:
# Process in parallel
tasks = []
for path in video_paths:
task = self._analyze_single_video(path, model)
tasks.append(task)
analyses = await asyncio.gather(*tasks)
for path, analysis in zip(video_paths, analyses):
results[path] = analysis
else:
# Process sequentially
for path in video_paths:
results[path] = await self._analyze_single_video(path, model)
return ToolResult(
success=True,
content=[TextContent(text=json.dumps(results, indent=2))]
)
except Exception as e:
return ToolResult(
success=False,
error=ToolError(message=f"Batch analysis failed: {str(e)}")
)
@self.server.tool()
async def compare_movements(
analysis_id1: str,
analysis_id2: str
) -> ToolResult:
"""
Compare movement patterns between two analyzed videos.
Args:
analysis_id1: First analysis ID
analysis_id2: Second analysis ID
Returns:
Comparison of movement metrics
"""
try:
if analysis_id1 not in self.analysis_cache:
return ToolResult(
success=False,
error=ToolError(message=f"Analysis ID '{analysis_id1}' not found")
)
if analysis_id2 not in self.analysis_cache:
return ToolResult(
success=False,
error=ToolError(message=f"Analysis ID '{analysis_id2}' not found")
)
# Get analyses
analysis1 = self.analysis_cache[analysis_id1]["json_output"]
analysis2 = self.analysis_cache[analysis_id2]["json_output"]
# Compare metrics
comparison = self._compare_analyses(analysis1, analysis2)
return ToolResult(
success=True,
content=[TextContent(text=json.dumps(comparison, indent=2))]
)
except Exception as e:
return ToolResult(
success=False,
error=ToolError(message=f"Comparison failed: {str(e)}")
)
async def _download_video(self, url: str) -> str:
"""Download video from URL to temporary file"""
async with httpx.AsyncClient() as client:
response = await client.get(url)
response.raise_for_status()
# Save to temp file
filename = Path(urlparse(url).path).name or "video.mp4"
temp_path = os.path.join(self.temp_dir, filename)
async with aiofiles.open(temp_path, 'wb') as f:
await f.write(response.content)
return temp_path
async def _analyze_single_video(self, path: str, model: str) -> Dict[str, Any]:
"""Analyze a single video"""
try:
if path.startswith(('http://', 'https://')):
path = await self._download_video(path)
json_output, _ = await asyncio.to_thread(
self.analyzer.process_video,
path,
model=model,
enable_visualization=False
)
return {
"status": "success",
"analysis": json_output
}
except Exception as e:
return {
"status": "error",
"error": str(e)
}
def _generate_summary(self, analysis_data: Dict[str, Any]) -> str:
"""Generate human-readable summary from analysis data"""
summary_parts = []
# Video info
video_info = analysis_data.get("video_info", {})
summary_parts.append(f"Video Analysis Summary")
summary_parts.append(f"Duration: {video_info.get('duration_seconds', 0):.1f} seconds")
summary_parts.append(f"Resolution: {video_info.get('width', 0)}x{video_info.get('height', 0)}")
summary_parts.append("")
# Movement summary
movement_summary = analysis_data.get("movement_analysis", {}).get("summary", {})
# Direction analysis
direction_data = movement_summary.get("direction", {})
dominant_direction = direction_data.get("dominant", "unknown")
summary_parts.append(f"Dominant Movement Direction: {dominant_direction}")
# Intensity analysis
intensity_data = movement_summary.get("intensity", {})
dominant_intensity = intensity_data.get("dominant", "unknown")
summary_parts.append(f"Movement Intensity: {dominant_intensity}")
# Speed analysis
speed_data = movement_summary.get("speed", {})
dominant_speed = speed_data.get("dominant", "unknown")
summary_parts.append(f"Movement Speed: {dominant_speed}")
# Segments
segments = movement_summary.get("movement_segments", [])
if segments:
summary_parts.append(f"\nMovement Segments: {len(segments)}")
for i, segment in enumerate(segments[:3]): # Show first 3
start_time = segment.get("start_time", 0)
end_time = segment.get("end_time", 0)
movement_type = segment.get("movement_type", "unknown")
summary_parts.append(f" Segment {i+1}: {movement_type} ({start_time:.1f}s - {end_time:.1f}s)")
return "\n".join(summary_parts)
def _compare_analyses(self, analysis1: Dict, analysis2: Dict) -> Dict[str, Any]:
"""Compare two movement analyses"""
comparison = {
"video1_info": analysis1.get("video_info", {}),
"video2_info": analysis2.get("video_info", {}),
"metric_comparison": {}
}
# Compare summaries
summary1 = analysis1.get("movement_analysis", {}).get("summary", {})
summary2 = analysis2.get("movement_analysis", {}).get("summary", {})
# Compare directions
dir1 = summary1.get("direction", {})
dir2 = summary2.get("direction", {})
comparison["metric_comparison"]["direction"] = {
"video1_dominant": dir1.get("dominant", "unknown"),
"video2_dominant": dir2.get("dominant", "unknown"),
"match": dir1.get("dominant") == dir2.get("dominant")
}
# Compare intensity
int1 = summary1.get("intensity", {})
int2 = summary2.get("intensity", {})
comparison["metric_comparison"]["intensity"] = {
"video1_dominant": int1.get("dominant", "unknown"),
"video2_dominant": int2.get("dominant", "unknown"),
"match": int1.get("dominant") == int2.get("dominant")
}
# Compare speed
speed1 = summary1.get("speed", {})
speed2 = summary2.get("speed", {})
comparison["metric_comparison"]["speed"] = {
"video1_dominant": speed1.get("dominant", "unknown"),
"video2_dominant": speed2.get("dominant", "unknown"),
"match": speed1.get("dominant") == speed2.get("dominant")
}
return comparison
async def run(self):
"""Run the MCP server"""
async with stdio_server() as (read_stream, write_stream):
await self.server.run(read_stream, write_stream)
async def main():
"""Main entry point"""
server = LabanMCPServer()
await server.run()
if __name__ == "__main__":
asyncio.run(main())