# OpenBench: Privacy-Preserving Benchmarking (MVP)

- This repository contains a minimal, working MVP of the OpenBench platform focused on:
- An offline-first KPI data model (Revenue, COGS, inventory turns, lead time, CAC, LTV).
- A simple, privacy-preserving aggregation primitive (Laplace-noise-enabled) for anonymized benchmarking.
- A lightweight Python packaging setup compatible with pytest-based tests and python -m build packaging checks.
- A minimal data-contract system (DataContract) and an in-memory contract registry for versioned payload schemas.
- A pluggable adapter framework with sample POSAdapter and ERPAdapter to bootstrap data ingestion.

How to run
- Install dependencies and run tests: `bash test.sh`
- The MVP stores KPI records locally in a JSONL file under the package data directory.

Project layout (high-level)
- openbench_privacy_preserving_benchmarkin/core.py: Core data model and analytics primitives.
- __init__.py: Re-exports core primitives for simple imports.
- test.sh: Quick test runner that also builds the distribution.
- AGENTS.md: Swarm agent guidelines describing architecture and testing commands.
- pyproject.toml/setup.py: Packaging metadata to satisfy python packaging checks.

This is a deliberately minimal MVP intended to demonstrate architecture and testing practices.