Metadata-Version: 2.4
Name: metanoia
Version: 0.1.0
Summary: Epistemic memory for AI agents — belief revision, provenance, and compound decay.
Project-URL: Repository, https://github.com/metanoia/metanoia
License: MIT
Keywords: agents,belief-revision,epistemic,llm,memory
Classifier: Development Status :: 1 - Planning
Classifier: Intended Audience :: Developers
Classifier: Intended Audience :: Science/Research
Classifier: License :: OSI Approved :: MIT License
Classifier: Programming Language :: Python :: 3
Classifier: Topic :: Scientific/Engineering :: Artificial Intelligence
Requires-Python: >=3.10
Description-Content-Type: text/markdown

# metanoia

> Epistemic memory for AI agents.

Memory systems for AI agents are epistemically inert — a corrected fact and an original fact have equal weight; a stale preference and a fresh one are equivalent. Metanoia fixes this.

Each memory carries **provenance** (how it was formed), **epistemic confidence** (how much to trust it), and participates in **compound decay** (time + contradiction events). Retrieval is weighted by epistemic quality, not just semantic similarity.

**Status:** Pre-release. Implementation in progress.

Built on [epistemic](https://pypi.org/project/epistemic/) — the provenance scoring and compound decay core.
