Metadata-Version: 2.4
Name: nomark-engine
Version: 0.1.0
Summary: NOMARK Engine — open-core agent outcome quality resolver
Project-URL: Homepage, https://github.com/nomark-dev/nomark
Project-URL: Repository, https://github.com/nomark-dev/nomark
Author: Reece Frazier
License-Expression: Apache-2.0
Keywords: agent,ai,intent,nomark,preferences,quality,trust
Classifier: Development Status :: 3 - Alpha
Classifier: Intended Audience :: Developers
Classifier: License :: OSI Approved :: Apache Software License
Classifier: Programming Language :: Python :: 3
Classifier: Programming Language :: Python :: 3.10
Classifier: Programming Language :: Python :: 3.11
Classifier: Programming Language :: Python :: 3.12
Classifier: Programming Language :: Python :: 3.13
Classifier: Topic :: Software Development :: Libraries
Requires-Python: >=3.10
Requires-Dist: pydantic<3,>=2.0
Description-Content-Type: text/markdown

# nomark-engine

Open-core agent outcome quality resolver. Understands what a human means from incomplete input by learning preferences across sessions and platforms.

## Install

```bash
pip install nomark-engine
```

## Quick start

```python
from nomark_engine import create_resolver, parse_ledger, ResolverConfig

# Parse a NOMARK ledger
entries = parse_ledger(open("nomark-ledger.jsonl").read())

# Create resolver
resolver = create_resolver(ResolverConfig(entries=entries))

# Resolve all preference dimensions
result = resolver.resolve_all()
for dim, res in result.dimensions.items():
    if res.winner:
        print(f"{dim}: {res.winner.pref.target} (score: {res.winner.score})")

# Resolve intent from natural language
result = resolver.resolve_input("make it shorter")
for match in result.meaning_maps:
    print(f"Matched: {match.trigger} -> {match.intent}")
```

## Modules

- **Schema** — Pydantic v2 models for all signal types (pref, map, asn, meta, rub)
- **Classifier** — Input tier classification (pass-through, routing, extraction)
- **Resolver** — MEE weighted scoring with scope matching and instability detection
- **Ledger** — JSONL parser/writer with capacity constraints
- **Decay** — Time-based decay with contradiction acceleration
- **Utility** — Multi-factor utility scoring and capacity-bounded pruning

## License

Apache 2.0
