embeddoor 0.1.0

A browser-based embedding visualization and analysis tool

Embeddoor

A browser-based tool for embedding visualization and analysis.

Features

• Dual-panel interface: 2D/3D plots on the left, custom visualizations (tables, images, word clouds) on the right
• Interactive data exploration: Load CSV files, visualize tabular data, and plot 2-3 numerical columns
• Advanced plot controls: Configure hue, size, and shape based on data columns
• Lasso selection: Select data points interactively and store selections in the dataframe
• Correlation analysis: Visualize pairwise correlations with Pearson, Spearman, or Kendall methods
• Heatmap visualizations: View data as heatmaps from embeddings or numeric columns
• Modular embedding framework: Create embeddings using HuggingFace, OpenAI, Gemini, and custom models
• Dimensionality reduction: Apply PCA, t-SNE, and UMAP to high-dimensional embeddings
• Data persistence: Save and load data in Parquet format

Installation

Development Installation

git clone https://github.com/haesleinhuepf/embeddoor.git
cd embeddoor
pip install -e .[dev,embeddings]

Quick Start

Launch the application:

embeddoor

This will start the server and open your default browser to http://localhost:5000.

Workflow

1. Load Data: Use File → Open to load a CSV file
2. Visualize: View tabular data in the right panel, plot numerical columns in the left panel
3. Customize Plot: Select hue, size, and shape attributes for data points
4. Select Points: Use the lasso tool to select data points (stored as a new column)
5. Create Embeddings: Embedding → Create Embedding to generate embeddings from text/image columns
6. Reduce Dimensions: Dimensionality Reduction → Apply PCA/t-SNE/UMAP to embeddings
7. Save: File → Save to export data as Parquet

License

MIT License

Contributing

Contributions are welcome! Please feel free to submit a Pull Request.