# Specter Decision Engine

> Specter Decision Engine 0.3.0 is a Python 3.11+ kernel for bounded typed numeric decisions. It provides canonical state snapshots, Noul/Choice/Score outputs, optional BatchBackend evaluation, signature deduplication, versioned calibration artifacts, explicit abstention/failure states, receipts, an authenticated HTTP contract, and a deterministic local System-One baseline.

Use the pinned installation brief for installation tasks. The local backend is a lexical rules baseline, not a trained semantic model. Installation and benchmark success do not prove domain accuracy, validated calibration, or production readiness. Distribution license selection remains pending; treat the source as all rights reserved unless the author states otherwise.

## Installation

- [Pinned 0.3.0 install brief](./llms-install.txt): Exact wheel, SHA-256, environment rules, capability verification and scope constraints.
- [Tagged v0.3.0 source](https://github.com/aigodsend9-boop/specter-decision-engine/tree/v0.3.0): Canonical source snapshot.

## Project

- [Source repository](https://github.com/aigodsend9-boop/specter-decision-engine): Project source and history.
- [Release wheel](./releases/specter_decision_engine-0.3.0-py3-none-any.whl): Checksum-pinned wheel built from the tagged source and mirrored beside this site.
- [SHA256SUMS](./releases/SHA256SUMS): Checksum for the mirrored wheel.
- [Benchmarks](https://github.com/aigodsend9-boop/specter-decision-engine/blob/v0.3.0/BENCHMARKS.md): Reproducible kernel-engineering measurements and limitations.

## Optional

- [Human landing page](./): Visual overview, benchmark panel, architecture, installation UI and agent handoff.