Metadata-Version: 2.4
Name: epistemic
Version: 0.1.0
Summary: Provenance scoring, compound decay, and confidence-weighted retrieval — the epistemic core.
Project-URL: Repository, https://github.com/metanoia/epistemic
License: MIT
Keywords: belief,confidence,epistemic,llm,provenance,retrieval
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

# epistemic

> Provenance scoring, compound decay, and confidence-weighted retrieval.

The domain-agnostic epistemic core: a retrieval layer that knows how much to trust what it remembers.

Every piece of stored knowledge carries a **provenance** (user-stated, agent-inferred, tool-observed, externally sourced), an **epistemic confidence** score, and a **compound decay** function that responds to both time and contradiction events. Retrieval is epistemic inference, not nearest-neighbor search.

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

Used by [metanoia](https://pypi.org/project/metanoia/) for agent memory. Designed for extension to legal, clinical, and code intelligence domains.
