constraint-theory 0.3.0

Pure Python Constraint Theory — temporal constraints, Eisenstein lattices, adaptive tolerance, PLATO tiles, baton shards

Constraint Theory v0.3.0

Pure Python constraint satisfaction toolkit — temporal constraints, Eisenstein lattices, adaptive tolerance, PLATO tiles, and baton shards.

pip install constraint-theory

Or for development:

git clone https://github.com/SuperInstance/constraint-theory-py
cd constraint-theory-py
pip install -e ".[dev]"
pytest

What's New in v0.3.0

1. Eisenstein Lattice Operations (constraint_theory.eisenstein)

Snap any 2-D point to the nearest A₂ lattice point with a guaranteed worst-case error of ~0.577 (covering radius). Includes full arithmetic on Eisenstein integers, Weyl chamber classification, dodecet encoding (12-bit compressed snap metadata), and norm computation.

from constraint_theory.eisenstein import snap, A2Point, encode, decode

pt = snap(0.5, 0.3)           # A2Point(a, b)
d = encode(1.2, 0.8)          # 12-bit Dodecet
err_lvl, angle, ch, safe = decode(d)

2. Temporal Constraints (constraint_theory.temporal)

Constraint propagation with exponential time decay. The deadband funnel narrows over time; anomaly spikes trigger re-widening. Model includes chirality (Weyl chamber commitment), prediction, and precision energy accumulation.

from constraint_theory.temporal import TemporalAgent

agent = TemporalAgent(decay_rate=0.8)
for step in range(100):
    update = agent.observe(x, y)
    if update.is_anomaly:
        print("Anomaly at step", step)

3. Adaptive Tolerance (constraint_theory.adaptive)

Formula ε(c) = min(k/c, ε_max) — as manifold curvature grows, snapping precision tightens. Includes region classification (far → approaching → near → critical → singular).

from constraint_theory.adaptive import AdaptiveTolerance

tol = AdaptiveTolerance(k=0.5)
eps_at_high_curve = tol(100.0)  # 0.005

4. PLATO Tile Interface (constraint_theory.plato)

Domain-scored knowledge tiles with relevance decay, reliability tracking, and composite scoring. Includes a lightweight in-memory store for prototyping.

from constraint_theory.plato import PlatoTile, PlatoTileStore

tile = PlatoTile(id="ct.001", domain="constraint-theory.eisenstein")
store = PlatoTileStore()
store.put(tile)

5. Baton Shard (constraint_theory.baton)

Split a context dict into three shards: artifacts, reasoning, blockers. Includes integrity hashing, JSON serialisation, and structural diffs.

from constraint_theory.baton import BatonShard, split_context

ctx = {"version": "1", "artifacts": {"data.txt": "..."}}
shard = split_context(ctx)
shard.add_blocker("Missing edge case")

Module Reference

Module	Classes / Functions	Tests
eisenstein	snap, A2Point, Dodecet, encode, decode, norm_sq, classify_chamber	36
temporal	TemporalAgent, snap_to_eisenstein, deadband_funnel	16
adaptive	AdaptiveTolerance, adaptive_epsilon, classify_region	12
plato	PlatoTile, PlatoTileStore	20
baton	BatonShard, split_context, merge_shards, diff_shards	18

Total: 100+ tests covering all modules.

Related Projects

Repo	Connection
eisenstein-embed	5-layer matching cascade using Eisenstein lattice arithmetic and adaptive tolerance
tensor-spline	SplineLinear compression built on Eisenstein lattices
plato-training	Micro model training with PLATO tile protocol

License

MIT