ELID

Embedding Locality IDentifier

Encode high-dimensional embedding vectors into compact, sortable strings. Similar embeddings produce lexicographically similar ELIDs, enabling vector search using standard string indexes. Plus: blazing-fast string similarity algorithms. All powered by Rust compiled to WebAssembly -- every computation runs entirely in your browser.

Rust + WASM Client-Side Only Zero Dependencies Multi-Language Bindings
@elid/core @elid/text @elid/image @elid/both
1

Embedding Encoding

Convert vector embeddings to sortable string identifiers

Encoding Profiles

Profile Use Case Reversible Output Size
Mini128 Fast similarity via Hamming distance No 26 chars
Morton10x10 Z-order curve, good for range queries No ~20 chars
Hilbert10x10 Best locality preservation No ~20 chars
Lossless Exact embedding storage and retrieval Yes ~dims x 6 chars
Compressed Tunable size/fidelity trade-off Yes Variable
MaxLength Fit in database column constraints Yes User-defined
T

Text Embedding

Generate 256-dim embeddings using Model2Vec (potion-base-8M) — runs entirely in your browser via WASM + ONNX

Text embedding requires WASM built with models-text feature.
2

Cross-Dimensional Comparison

Compare embeddings from different models with different dimensions

3

Works With Any Embedding

Images, audio, code — if you can embed it, ELID can index it

Vector Search Without Vector Stores

The Problem

Semantic search requires vector databases (Pinecone, Milvus, Weaviate) that add infrastructure complexity and cost. You cannot use standard SQL databases efficiently for similarity search because vectors don't have natural ordering.

The ELID Solution

ELID encodes embeddings into strings where lexicographic order approximates semantic similarity. Store ELIDs in any database with a B-tree index (PostgreSQL, SQLite, DynamoDB) and use standard string operations for similarity search.

Example Usage

-- Store with standard index
CREATE INDEX idx_elid ON documents(elid);

-- Find similar documents
SELECT * FROM documents
WHERE elid BETWEEN 'abc...' AND 'abd...'
ORDER BY elid
LIMIT 10;

Works with any database that supports string indexing. No special extensions required.

String Similarity Tools
4

Algorithm Comparison

Compare two strings across every algorithm in real time

Examples:
6

SimHash Visualizer

See how similar strings produce similar 64-bit fingerprints

Click on two entries to compare their bit patterns. Differing bits are highlighted in red.