Skip to content

README.md

Latest commit

 

History

History
25 lines (16 loc) · 1016 Bytes

README.md

File metadata and controls

25 lines (16 loc) · 1016 Bytes

Embedding Witch

A browser-based tool for generating text embeddings using local HuggingFace models. No server required - everything runs client-side.

Features

  • Local models — Browse and download compatible embedding models from HuggingFace
  • Flexible input — Paste text directly or upload a file
  • Split modes — Embed by lines, paragraphs, or whole text
  • Export — Download embeddings as JSON, CSV, TSV, or NPY

Model Size Limit

Only models with ≤100M parameters (and known parameter counts) are shown in the dropdown. Larger models are excluded because they can cause out-of-memory crashes and unacceptably slow inference in the browser. These models can be optionally restored.

Testing Models

Run the compatibility test to verify which models can actually download and produce embeddings:

npm install
node test-models.js

Results are saved to test-results.json. Each model is tested by loading it and embedding "Hello world" with a 120-second timeout.