ArrowSpace Skills

Portable skill definitions for working with ArrowSpace — a spectral vector search library that goes beyond cosine similarity by incorporating graph Laplacian eigenstructure.

ArrowSpace adds a spectral dimension to vector similarity, enabling search that considers both semantic content and structural role in the dataset.

Quick Start

pip install arrowspace

Then:

from arrowspace import ArrowSpaceBuilder
import numpy as np

items = np.array([[0.1, 0.2, 0.3], [0.0, 0.5, 0.1], [0.9, 0.1, 0.0]], dtype=np.float64)
graph_params = {"eps": 1.0, "k": 6, "topk": 3, "p": 2.0, "sigma": 1.0}
aspace, gl = ArrowSpaceBuilder().build(graph_params, items)

query = np.array([0.05, 0.2, 0.25], dtype=np.float64)
hits = aspace.search(query, gl, tau=1.0)
print(hits)  # [(0, 0.99), (1, 0.76), (2, 0.22)]

Resources

Skills

This repo contains skill definitions that teach AI agents how to use ArrowSpace:

Adding to an AI Agent

OpenCode — symlink or copy SKILL.md into .opencode/skills/:

cp SKILL.md .opencode/skills/arrowspace/SKILL.md

Claude Code — reference in CLAUDE.md or add to project instructions.

Copilot / Cursor — include the relevant skills/*.md content in your custom instructions.

Any agent — point it at this repo: "You have access to the ArrowSpace skills at skills/."

Python Package

Install the skill utilities directly:

pip install arrowspace-skills

from arrowspace_skills import suggest_params, build_index, tune_tau, item_lambdas

See arrowspace_skills/ for reusable helper functions.

License

Apache 2.0 — same as arrowspace-rs.