Skip to content

README.md

Latest commit

 

History

History
142 lines (107 loc) · 3.38 KB

README.md

File metadata and controls

142 lines (107 loc) · 3.38 KB
title OpenPIV MCP Server
emoji 🌊
colorFrom blue
colorTo purple
sdk docker
sdk_version edge
pinned false
license mit

OpenPIV MCP Server

Particle Image Velocimetry (PIV) analysis via MCP protocol.

Features

  • compute_piv: Compute velocity fields from image pairs
  • create_quiver_plot: Generate vector field visualizations
  • Web UI: Interactive Gradio interface for uploading images and viewing results
  • MCP Protocol: Connect to Claude Desktop, Cursor, Windsurf, and other MCP clients

Quick Start

Option 1: Local MCP Server (Recommended for Development)

Run the MCP server locally using stdio transport:

# Clone and install
git clone https://huggingface.co/spaces/alexliberzon/openpiv-mcp
cd openpiv-mcp
pip install -r requirements.txt

# Run the MCP server
python src/openpiv_mcp.py

Claude Desktop config (~/Library/Application Support/Claude/claude_desktop_config.json):

{
  "mcpServers": {
    "openpiv": {
      "command": "python",
      "args": ["/absolute/path/to/openpiv-mcp/src/openpiv_mcp.py"]
    }
  }
}

Option 2: Hugging Face Spaces (Remote HTTP)

The server is deployed on Hugging Face Spaces with both web UI and MCP endpoint:

MCP Client Config:

{
  "mcpServers": {
    "openpiv": {
      "url": "https://alexliberzon-openpiv-mcp.hf.space/gradio_api/mcp/"
    }
  }
}

Note: When using the remote MCP server, images must be provided as:

  • HTTP/HTTPS URLs (e.g., https://example.com/image.png)
  • Base64 data URLs (e.g., data:image/png;base64,iVBOR...)

Option 3: Web UI Only

Visit https://alexliberzon-openpiv-mcp.hf.space to use the interactive web interface:

  1. Upload two consecutive frames (Frame A and Frame B)
  2. Adjust parameters (window size, overlap, time delay)
  3. Click "Compute PIV" to get velocity field data
  4. Use the "Quiver Plot" tab to visualize results

API Reference

compute_piv

Compute Particle Image Velocimetry velocity field from two images.

Parameters:

  • image_a (str): Image URL, base64 data URL, or file path (local only)
  • image_b (str): Image URL, base64 data URL, or file path (local only)
  • window_size (int): Interrogation window size (default: 32)
  • overlap (int): Overlap between windows (default: 16)
  • dt (float): Time delay between frames (default: 1.0)

Returns: Summary statistics and CSV file path

create_quiver_plot

Create a quiver (vector field) plot from PIV results.

Parameters:

  • csv_file (str): Path to PIV results CSV file
  • title (str): Plot title (default: "PIV Velocity Field")
  • scale (int): Quiver scale factor (default: 50)
  • cmap (str): Colormap (default: "viridis")

Returns: PIL Image of the quiver plot

Example Output

**PIV computation successful!**

**Summary Statistics:**
- Total vectors computed: 660
- Valid vectors (s2n>1): 580
- Mean U velocity: -0.0814
- Max U velocity: 1.7142
- Max V velocity: 6.9067

Local Testing

# Test with demo images
python -c "
from openpiv_client import compute_piv, create_quiver_plot
import asyncio

async def test():
    result = await compute_piv(
        image_a_path='demo/test1/exp1_001_a.bmp',
        image_b_path='demo/test1/exp1_001_b.bmp'
    )
    print(result)

asyncio.run(test())
"

License

MIT License