test(goldtether): RED #18 bootstrap/prune (proof gate + principal-axis mode)
Add GoldPromotionProof, expand promote/prune signatures, promotion_eligible stub, and behavioral suite for residual/closure gates, require_proof, principal_axes prune, and SERVE containment. Intentionally incomplete: principal_axes and eligibility raise NotImplementedError; dirty promote still admitted — RED until GREEN.
This commit is contained in:
parent
9ab4cca821
commit
11e7b71d57
3 changed files with 303 additions and 10 deletions
|
|
@ -29,6 +29,7 @@ from core.physics.goldtether import (
|
|||
AutonomyBand,
|
||||
AutonomyDecision,
|
||||
CoherenceResidual,
|
||||
GoldPromotionProof,
|
||||
GoldTetherMonitor,
|
||||
OperatingMode,
|
||||
coherence_residual,
|
||||
|
|
@ -83,7 +84,7 @@ __all__ = [
|
|||
"IdentityManifold", "IdentityCheck", "IdentityScore", "CharacterProfile",
|
||||
"PromotionDecision", "VaultPromotionPolicy",
|
||||
"AutonomyBand", "AutonomyDecision", "CoherenceResidual",
|
||||
"GoldTetherMonitor", "OperatingMode", "coherence_residual",
|
||||
"GoldPromotionProof", "GoldTetherMonitor", "OperatingMode", "coherence_residual",
|
||||
"AxisClassification", "CartanIwasawaFactors", "ConformalProcrustesResult",
|
||||
"PrincipalAxis", "SignatureAwarePCAResult",
|
||||
"cartan_iwasawa_extract", "cartan_iwasawa_factorize",
|
||||
|
|
|
|||
|
|
@ -24,7 +24,7 @@ from __future__ import annotations
|
|||
|
||||
from dataclasses import dataclass, field
|
||||
from enum import Enum
|
||||
from typing import Any, Optional, Tuple
|
||||
from typing import Any, Literal, Optional, Tuple
|
||||
|
||||
import numpy as np
|
||||
|
||||
|
|
@ -39,6 +39,23 @@ _TELEMETRY_SCHEMA = "goldtether_coherence_v2" # v2: dropped vacuous grade-5 cha
|
|||
_E4_IDX = 4
|
||||
_E5_IDX = 5
|
||||
|
||||
PruneMode = Literal["fifo", "principal_axes"]
|
||||
|
||||
|
||||
@dataclass(frozen=True, slots=True)
|
||||
class GoldPromotionProof:
|
||||
"""Caller-supplied proof for replay-verified promotion into 𝓘_gold (ADR-0092).
|
||||
|
||||
Physics never signs reviews. The review surface constructs this payload and
|
||||
passes ``authorized=True`` only after external verification. ``replay_hash``
|
||||
is opaque to GoldTether (determinism pin for the caller).
|
||||
"""
|
||||
|
||||
residual: float
|
||||
replay_hash: str
|
||||
reviewer_id: str
|
||||
closed: bool
|
||||
|
||||
|
||||
def _primal_gold_invariants() -> list:
|
||||
"""R&D-Revised §5 bootstrapping seeds: the identity versor and the two
|
||||
|
|
@ -273,21 +290,63 @@ class GoldTetherMonitor:
|
|||
alpha = self.alpha_constraint(F, mode=mode)
|
||||
return self.supervised_blend(v_self, v_constraint, alpha)
|
||||
|
||||
def promote_gold_invariant(self, F: np.ndarray, *, authorized: bool = False) -> None:
|
||||
"""Add a state versor to 𝓘_gold. CALLER-GATED: the ADR-0092 signed /
|
||||
replay-verified promotion happens in the caller; this refuses to
|
||||
self-authorize (one-mutation-path discipline). The full replay-verified
|
||||
promotion pipeline + principal-axis decay are deferred (issue #18 follow-up).
|
||||
def promotion_eligible(self, F: np.ndarray) -> bool:
|
||||
"""True iff F is closed and harmonized residual is at/below ε_drift.
|
||||
|
||||
Bootstrap gate (R&D §5): only audit-grade coherent states are candidates
|
||||
for 𝓘_gold. Does not authorize promotion by itself.
|
||||
"""
|
||||
# RED until #18 GREEN: live residual/closure check not yet wired as eligibility.
|
||||
raise NotImplementedError(
|
||||
"promotion_eligible: issue #18 bootstrap eligibility not implemented"
|
||||
)
|
||||
|
||||
def promote_gold_invariant(
|
||||
self,
|
||||
F: np.ndarray,
|
||||
*,
|
||||
authorized: bool = False,
|
||||
proof: Optional[GoldPromotionProof] = None,
|
||||
require_proof: bool = False,
|
||||
) -> None:
|
||||
"""Add a state versor to 𝓘_gold. CALLER-GATED (ADR-0092).
|
||||
|
||||
- Without ``authorized=True``: refuse (proposal-only; proof alone is insufficient).
|
||||
- With authorize: refuse non-closed or high-residual F (live check, not proof trust).
|
||||
- ``require_proof=True``: refuse if ``proof`` is missing.
|
||||
- Physics never self-signs reviews; ``proof`` is caller-supplied audit pin.
|
||||
|
||||
Partial (#24): authorize-only append. Full residual/proof gates are #18 GREEN.
|
||||
"""
|
||||
if not authorized:
|
||||
raise ValueError(
|
||||
"promote_gold_invariant requires explicit authorization (ADR-0092 gate)"
|
||||
)
|
||||
# Intentionally incomplete for RED: does not yet refuse dirty F / require proof.
|
||||
if require_proof and proof is None:
|
||||
# Minimal stub so require_proof tests can go red on residual/close paths first.
|
||||
raise ValueError("promote_gold_invariant requires proof when require_proof=True")
|
||||
self.gold_invariants.append(_as_mv(F).copy())
|
||||
|
||||
def prune_gold_invariants(self, max_size: int = 64) -> None:
|
||||
"""Bound 𝓘_gold (decay hook), always retaining the three primal seeds.
|
||||
Full principal-axis pruning (R&D-Revised §5) is deferred."""
|
||||
def prune_gold_invariants(
|
||||
self,
|
||||
max_size: int = 64,
|
||||
*,
|
||||
mode: PruneMode | str = "fifo",
|
||||
) -> None:
|
||||
"""Bound 𝓘_gold, always retaining the three primal seeds.
|
||||
|
||||
Modes:
|
||||
* ``fifo`` — keep primals + most recent (landed #24).
|
||||
* ``principal_axes`` — R&D §5 principal-axis decay (#18; RED until GREEN).
|
||||
"""
|
||||
mode_s = str(mode)
|
||||
if mode_s not in ("fifo", "principal_axes"):
|
||||
raise ValueError(f"prune_gold_invariants unknown mode: {mode_s!r}")
|
||||
if mode_s == "principal_axes":
|
||||
raise NotImplementedError(
|
||||
"prune_gold_invariants(mode='principal_axes'): issue #18 not implemented"
|
||||
)
|
||||
max_size = max(3, int(max_size))
|
||||
if len(self.gold_invariants) > max_size:
|
||||
primal = self.gold_invariants[:3]
|
||||
|
|
|
|||
233
tests/test_adr_0238_goldtether_bootstrap_prune.py
Normal file
233
tests/test_adr_0238_goldtether_bootstrap_prune.py
Normal file
|
|
@ -0,0 +1,233 @@
|
|||
"""ADR-0238 / R&D §5 — GoldTether 𝓘_gold bootstrap + principal-axis prune (#18).
|
||||
|
||||
Residual+α already landed (#24). Remaining fidelity:
|
||||
* promote with proof payload (closed + residual gate + optional replay hash)
|
||||
* still proposal-only unless authorized (ADR-0092 gate)
|
||||
* principal-axis prune retains primals, drops low-eigen members
|
||||
* retained members stay closed (versor_condition < 1e-6)
|
||||
|
||||
See docs/research/third-door-blueprint-fidelity.md finding #4 / issue #18.
|
||||
"""
|
||||
|
||||
from __future__ import annotations
|
||||
|
||||
import hashlib
|
||||
|
||||
import numpy as np
|
||||
import pytest
|
||||
|
||||
from algebra.rotor import make_rotor_from_angle
|
||||
from algebra.versor import versor_condition
|
||||
from core.physics.goldtether import (
|
||||
GoldPromotionProof,
|
||||
GoldTetherMonitor,
|
||||
OperatingMode,
|
||||
)
|
||||
|
||||
|
||||
def _id() -> np.ndarray:
|
||||
v = np.zeros(32, dtype=np.float64)
|
||||
v[0] = 1.0
|
||||
return v
|
||||
|
||||
|
||||
def _closed_rotor(angle: float = 0.3, plane: int = 6) -> np.ndarray:
|
||||
return make_rotor_from_angle(angle, bivector_idx=plane)
|
||||
|
||||
|
||||
def _dirty() -> np.ndarray:
|
||||
v = np.zeros(32, dtype=np.float64)
|
||||
v[0] = 0.5
|
||||
v[1] = 0.5
|
||||
return v
|
||||
|
||||
|
||||
def _proof_for(F: np.ndarray, *, residual: float | None = None) -> GoldPromotionProof:
|
||||
m = GoldTetherMonitor()
|
||||
r = float(m.goldtether_residual(F) if residual is None else residual)
|
||||
payload = np.asarray(F, dtype=np.float64).tobytes() + f"{r:.12g}".encode()
|
||||
return GoldPromotionProof(
|
||||
residual=r,
|
||||
replay_hash=hashlib.sha256(payload).hexdigest(),
|
||||
reviewer_id="test-reviewer",
|
||||
closed=versor_condition(F) < 1e-6,
|
||||
)
|
||||
|
||||
|
||||
# --- promote: ADR-0092 gate + proof discipline ---------------------------------
|
||||
|
||||
|
||||
def test_promote_refuses_without_authorization_even_with_proof():
|
||||
"""Proposal-only: proof alone must not mutate 𝓘_gold."""
|
||||
m = GoldTetherMonitor()
|
||||
F = _closed_rotor(0.2)
|
||||
proof = _proof_for(F)
|
||||
with pytest.raises(ValueError, match="authorization|ADR-0092|authorized"):
|
||||
m.promote_gold_invariant(F, proof=proof)
|
||||
assert len(m.gold_invariants) == 3
|
||||
|
||||
|
||||
def test_promote_refuses_without_authorization():
|
||||
m = GoldTetherMonitor()
|
||||
with pytest.raises(ValueError, match="authorization|ADR-0092|authorized"):
|
||||
m.promote_gold_invariant(_closed_rotor(0.3))
|
||||
assert len(m.gold_invariants) == 3
|
||||
|
||||
|
||||
def test_promote_refuses_non_closed_even_when_authorized():
|
||||
m = GoldTetherMonitor()
|
||||
dirty = _dirty()
|
||||
with pytest.raises(ValueError, match="closed|versor|closure"):
|
||||
m.promote_gold_invariant(dirty, authorized=True, proof=_proof_for(dirty, residual=0.0))
|
||||
assert len(m.gold_invariants) == 3
|
||||
|
||||
|
||||
def test_promote_refuses_high_residual_even_when_authorized():
|
||||
"""Drift-loud states must not enter 𝓘_gold even under explicit authorize."""
|
||||
m = GoldTetherMonitor()
|
||||
dirty = _dirty()
|
||||
# proof claims residual 0 but live residual is large — refuse
|
||||
with pytest.raises(ValueError, match="residual|ε|epsilon|drift"):
|
||||
m.promote_gold_invariant(
|
||||
dirty,
|
||||
authorized=True,
|
||||
proof=GoldPromotionProof(
|
||||
residual=0.0,
|
||||
replay_hash="0" * 64,
|
||||
reviewer_id="test",
|
||||
closed=False,
|
||||
),
|
||||
)
|
||||
assert len(m.gold_invariants) == 3
|
||||
|
||||
|
||||
def test_promote_requires_proof_when_require_proof_true():
|
||||
m = GoldTetherMonitor()
|
||||
F = _closed_rotor(0.15)
|
||||
with pytest.raises(ValueError, match="proof"):
|
||||
m.promote_gold_invariant(F, authorized=True, require_proof=True)
|
||||
assert len(m.gold_invariants) == 3
|
||||
|
||||
|
||||
def test_promote_authorized_closed_low_residual_with_proof_appends():
|
||||
m = GoldTetherMonitor()
|
||||
F = _id() # on-seed, residual 0
|
||||
proof = _proof_for(F)
|
||||
m.promote_gold_invariant(F, authorized=True, proof=proof, require_proof=True)
|
||||
assert len(m.gold_invariants) == 4
|
||||
assert np.allclose(m.gold_invariants[-1], F)
|
||||
|
||||
|
||||
def test_promote_determinism_same_proof_payload():
|
||||
F = _closed_rotor(0.4)
|
||||
p1 = _proof_for(F)
|
||||
p2 = _proof_for(F)
|
||||
assert p1.replay_hash == p2.replay_hash
|
||||
assert p1.residual == p2.residual
|
||||
|
||||
|
||||
# --- prune: principal-axis mode -----------------------------------------------
|
||||
|
||||
|
||||
def test_prune_fifo_mode_retains_primals():
|
||||
"""Backward-compatible default: size bound keeps three primals."""
|
||||
m = GoldTetherMonitor()
|
||||
for i in range(40):
|
||||
m.promote_gold_invariant(_closed_rotor(0.01 * i + 0.05), authorized=True)
|
||||
m.prune_gold_invariants(max_size=10, mode="fifo")
|
||||
assert len(m.gold_invariants) == 10
|
||||
assert m.gold_invariants[0][0] == 1.0
|
||||
assert m.gold_invariants[1][5] == 0.5
|
||||
assert m.gold_invariants[2][4] == 1.0 and m.gold_invariants[2][5] == 1.0
|
||||
|
||||
|
||||
def test_prune_principal_axes_retains_primals_and_bounds_size():
|
||||
m = GoldTetherMonitor()
|
||||
for i in range(30):
|
||||
# Stay on Euclidean rotation planes (6–8); boost planes need cosh form.
|
||||
m.promote_gold_invariant(
|
||||
_closed_rotor(0.05 * i + 0.1, plane=6 + (i % 3)),
|
||||
authorized=True,
|
||||
)
|
||||
n_before = len(m.gold_invariants)
|
||||
assert n_before > 10
|
||||
m.prune_gold_invariants(max_size=8, mode="principal_axes")
|
||||
assert len(m.gold_invariants) <= 8
|
||||
assert len(m.gold_invariants) >= 3
|
||||
# Three primal seeds always first and unchanged
|
||||
assert m.gold_invariants[0][0] == 1.0
|
||||
assert m.gold_invariants[1][5] == 0.5 and m.gold_invariants[1][4] == -0.5
|
||||
assert m.gold_invariants[2][4] == 1.0 and m.gold_invariants[2][5] == 1.0
|
||||
|
||||
|
||||
def test_prune_principal_axes_retained_members_are_closed():
|
||||
m = GoldTetherMonitor()
|
||||
for i in range(20):
|
||||
m.promote_gold_invariant(_closed_rotor(0.08 * i + 0.05), authorized=True)
|
||||
m.prune_gold_invariants(max_size=6, mode="principal_axes")
|
||||
for i, inv in enumerate(m.gold_invariants):
|
||||
# Primals n_o / n_inf are null (not unit versors); skip strict unit check
|
||||
if i == 0:
|
||||
assert versor_condition(inv) < 1e-6
|
||||
elif i >= 3:
|
||||
assert versor_condition(inv) < 1e-6, f"member[{i}] not closed"
|
||||
|
||||
|
||||
def test_prune_principal_axes_differs_from_fifo_on_ordered_promotions():
|
||||
"""PCA prune is not last-N FIFO: different retention among non-primals.
|
||||
|
||||
Promote a long chain; FIFO keeps the most recent, principal_axes ranks by
|
||||
energy/span. With enough spread, the retained non-primal sets must differ
|
||||
(or at least the API mode must be accepted and size-bounded).
|
||||
"""
|
||||
m_fifo = GoldTetherMonitor()
|
||||
m_pca = GoldTetherMonitor()
|
||||
for i in range(25):
|
||||
F = _closed_rotor(0.12 * i + 0.05, plane=6 + (i % 3))
|
||||
m_fifo.promote_gold_invariant(F, authorized=True)
|
||||
m_pca.promote_gold_invariant(F, authorized=True)
|
||||
m_fifo.prune_gold_invariants(max_size=7, mode="fifo")
|
||||
m_pca.prune_gold_invariants(max_size=7, mode="principal_axes")
|
||||
assert len(m_fifo.gold_invariants) == 7
|
||||
assert len(m_pca.gold_invariants) <= 7
|
||||
# Non-primal tails: PCA must not be a silent alias of FIFO forever.
|
||||
fifo_tail = [tuple(np.round(v, 8)) for v in m_fifo.gold_invariants[3:]]
|
||||
pca_tail = [tuple(np.round(v, 8)) for v in m_pca.gold_invariants[3:]]
|
||||
assert fifo_tail != pca_tail or len(pca_tail) < len(fifo_tail)
|
||||
|
||||
|
||||
def test_prune_unknown_mode_refused():
|
||||
m = GoldTetherMonitor()
|
||||
with pytest.raises(ValueError, match="mode"):
|
||||
m.prune_gold_invariants(max_size=5, mode="not_a_mode")
|
||||
|
||||
|
||||
def test_prune_refuses_stripping_below_primal_count():
|
||||
"""max_size < 3 must not drop primal seeds (clamp / refuse)."""
|
||||
m = GoldTetherMonitor()
|
||||
m.promote_gold_invariant(_closed_rotor(0.2), authorized=True)
|
||||
m.prune_gold_invariants(max_size=2, mode="principal_axes")
|
||||
assert len(m.gold_invariants) >= 3
|
||||
assert m.gold_invariants[0][0] == 1.0
|
||||
|
||||
|
||||
# --- eligibility helper (bootstrap pipeline surface) --------------------------
|
||||
|
||||
|
||||
def test_promotion_eligible_true_for_closed_on_seed():
|
||||
m = GoldTetherMonitor()
|
||||
assert m.promotion_eligible(_id()) is True
|
||||
|
||||
|
||||
def test_promotion_eligible_false_for_dirty():
|
||||
m = GoldTetherMonitor()
|
||||
assert m.promotion_eligible(_dirty()) is False
|
||||
|
||||
|
||||
def test_serve_mode_never_relaxes_via_gold_set_growth():
|
||||
"""Containment: growing 𝓘_gold does not make SERVE autonomous."""
|
||||
m = GoldTetherMonitor()
|
||||
m.autonomy = 1.0
|
||||
for i in range(5):
|
||||
m.promote_gold_invariant(_closed_rotor(0.05 * i), authorized=True)
|
||||
assert m.alpha_constraint(_id(), mode=OperatingMode.SERVE) == 1.0
|
||||
Loading…
Reference in a new issue