nomark-engine

The open-core preference resolution engine for AI agents. Python port of @nomark-ai/engine — identical resolution logic, Pydantic v2 models.

Install

pip install nomark-engine

Requires Python 3.10+.

Quick Start

from nomark_engine import create_resolver, parse_ledger, ResolverConfig

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

# Create a 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 five signal types: SigPref, SigMap, SigAsn, SigMeta, SigRub. Discriminated unions via LedgerEntry.

Ledger

JSONL parser and writer with per-type capacity constraints and token estimation.

Classifier

Input tier classification: pass-through, routing, or extraction.

Resolver

Five-factor weighted scoring engine:

• Specificity (0.30) — compound scope > single scope > global
• Evidence (0.25) — signal count normalized to 20
• Recency (0.20) — linear decay over 180 days
• Stability (0.15) — contradiction ratio penalty
• Portability (0.10) — cross-context usage

Unstable dimensions (winner score < 0.4) recommend asking instead of guessing.

Decay

Time-based weight decay with contradiction acceleration.

Utility

Multi-factor utility scoring and capacity-bounded pruning.

API Reference

Core

Types

Parity

This package produces identical results to @nomark-ai/engine for identical inputs. The test suites validate cross-language consistency.

License

Apache 2.0