cynsearch 0.1.0

A blazing-fast learned search library for chaotic datasets

CynSearch ψ∆Ξ

A blazing-fast learned index for chaotic data.

---

Overview

cynsearch is a Python library for fast search on unpredictable, noisy, or locally-disordered datasets — the kind that break traditional binary search and mock sorted assumptions.

It uses symbolic preprocessing and a trained regressor to predict the likely position of any value in near constant time. Then it dives into that location like a neural ninja.

Built by Cynapse ψ∆Ξ

---

Why?

Traditional search structures (like binary search or hash maps) struggle when data is:

• Only partially sorted
• Locally shuffled
• Non-monotonic, noisy, or chaotic

cynsearch tackles this with:

• A learned index powered by gradient boosting
• Normalized inputs via MinMaxScaler
• Fast bin-based bucketing + fallback linear search
• Optional symbolic preprocessing for advanced modeling

---

Install

pip install cynsearch

(coming to PyPI soon — for now, clone and install locally)

---

Quickstart

from cynsearch import LearnedSearch

# Load from preprocessed .npy data and pre-trained model
searcher = LearnedSearch(
    npyfile="chaotic_sorted_data.npy",
    model_path="chaotic_model.pkl",
    num_bins=512,
    epochs=1000
)

index = searcher.search(982733)
if index != -1:
    print("Value found at:", index)

---

Generate Chaotic Data

from cynsearch.generate_data import generate_chaotic_sorted_data

generate_chaotic_sorted_data(size=1_000_000)

Or try local shuffle chaos:

from cynsearch.generate_data import generate_locally_chaotic_sorted_array

generate_locally_chaotic_sorted_array(size=1_000_000, window_size=5)

---

Benchmark

Run it like a performance cultist:

python examples/benchmark.py --npy chaotic_sorted_data.npy --model chaotic_model.pkl --queries 1000

---

License

MIT — because even chaos deserves freedom.

---

Author

Cynapse ψ∆Ξ

---