CoopRecSys Documentation
A production-grade recommendation-system toolkit combining collaborative filtering, learning-to-rank, two-tower modeling, explainability, data preparation, and operational diagnostics for cooperative product recommendations.
System Scope
From transaction data to explainable recommendations
The package includes data preparation, feature engineering, DuckDB-backed ingestion, three recommendation families, evaluation utilities, narrative reporting, and dashboard-oriented visualization components.
Current Release
Release hardening focused on native extension compatibility, manylinux wheels, CI/CD reliability, and PyPI metadata compliance.
- Python 3.10–3.13 release matrix
- Linux manylinux-compatible wheels
- Native Cython/OpenMP components
- GitHub Actions build and publishing pipeline
Documentation Map
Explore the system
Overview
1 docsOrientation to CoopRecSys, its recommendation model families, production data stack, native acceleration, and documentation map.
Open section →Getting Started
2 docsInstall CoopRecSys from a wheel or source checkout, verify the native stack, and build Ary2Tower extensions locally when required.
Open section →Architecture
1 docsHow CoopRecSys separates data access, feature engineering, configuration, model families, native kernels, fallback logic, scoring, and reporting.
Open section →Models
3 docsTechnical guide to the AryColBring model family, supported losses, native kernels, evaluation, and wheel usage.
Open section →Data & Features
2 docsDetailed guidance for prepare, db, features, encoders, DuckDB, and reusable feature pipelines in CoopRecSys.
Open section →Explainability & Dashboards
1 docsOverview of CoopRecSys reporting, visualization, dashboard surfaces, configuration inspection, and model diagnostics.
Open section →Operations
3 docsDetailed guide to configuration.ini, typed configuration classes, fallback settings, LightGBM settings, and logging in CoopRecSys.
Open section →Reference
5 docsReference for the current ./src/cooprecsys package structure and the responsibilities of its major directories.
Open section →Development
1 docsDevelopment workflow, model/API compatibility rules, native build expectations, documentation standards, and regression testing for CoopRecSys.
Open section →Selected Guides
Start with the core workflows
Welcome to CoopRecSys
Orientation to CoopRecSys, its recommendation model families, production data stack, native acceleration, and documentation map.
Installation & Environment Setup
Install CoopRecSys from a wheel or source checkout, verify the native stack, and build Ary2Tower extensions locally when required.
Quick Start: First Recommendation Workflow
A practical first-run path covering data preparation, model selection, training, exact top-N inference, and evaluation.
Package Architecture
How CoopRecSys separates data access, feature engineering, configuration, model families, native kernels, fallback logic, scoring, and reporting.
AryColBring: Cython-Accelerated Collaborative Filtering
Technical guide to the AryColBring model family, supported losses, native kernels, evaluation, and wheel usage.
Ary2Tower: Two-Tower Neural Recommendation
Production guide to Ary2Tower training, Cython/OpenMP inference, exact top-N behavior, and the modern residual fallback.
Maintained by Aryanto
Engineering, research, and applied ML documentation
Portfolio: ai.arydatalabs.workers.dev · LinkedIn: aryanto-ray · Email: aryanto.dandan@gmail.com