dafit
f640dbdd65
- CLI: nyx-probe scan with --summary/--delta/--full flags - DriftProbe: training safety with Gini coefficient + Angular Drift - Vocabulary: 54 terms (30 nimmerverse + 24 German philosophical) - Sentinels: ANCHOR/BRIDGE/CANARY/TARGET monitoring system Key findings: - German philosophical terms: 37.5% depth≥2 hit rate (vs 3.3% nimmerverse) - Super Cluster validated: heart cross-lang sim = 1.000 - Isolated Zone confirmed: being EN↔DE sim = 0.195 - Gini signature: Philosophy ~0.5 (diffuse), Technical ~0.8 (sparse) 🤖 Generated with [Claude Code](https://claude.com/claude-code) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
98 lines
3.1 KiB
Python
98 lines
3.1 KiB
Python
"""
|
|
Result dataclasses for probing operations.
|
|
|
|
These structures capture what we learn about each term.
|
|
"""
|
|
from dataclasses import dataclass, field
|
|
from typing import List, Optional, Literal
|
|
from datetime import datetime
|
|
from enum import Enum
|
|
|
|
|
|
class EchoType(str, Enum):
|
|
"""Classification of echo probe responses."""
|
|
|
|
EXPANDS = "EXPANDS" # Real depth - adds new information
|
|
CONFIRMS = "CONFIRMS" # Shallow but solid - reinforces without adding
|
|
CIRCULAR = "CIRCULAR" # Surface only - returns to original term
|
|
DIVERGENT = "DIVERGENT" # Wrong direction - unrelated tangent
|
|
COLLAPSE = "COLLAPSE" # Nothing there - incoherent or empty
|
|
|
|
|
|
class ReadinessLevel(str, Enum):
|
|
"""Readiness classification for curriculum design."""
|
|
|
|
HIGH = "HIGH" # Ready for state machine / direct training
|
|
MEDIUM = "MEDIUM" # Needs scaffolding / bridging concepts
|
|
LOW = "LOW" # Requires foundational work first
|
|
|
|
|
|
@dataclass
|
|
class SurfaceProbeResult:
|
|
"""Result from a surface probe (single word → completions)."""
|
|
|
|
term: str
|
|
completions: List[str]
|
|
hit_eos_count: int # How many completions ended with EOS
|
|
avg_tokens: float # Average completion length
|
|
|
|
# Optional analysis
|
|
coherence_score: Optional[float] = None # 0-1, how related are completions
|
|
|
|
timestamp: datetime = field(default_factory=datetime.now)
|
|
|
|
|
|
@dataclass
|
|
class EchoProbeResult:
|
|
"""Result from an echo probe (iterative depth measurement)."""
|
|
|
|
term: str
|
|
rounds: int
|
|
chain: List[str] # The sequence of prompts/completions
|
|
echo_types: List[EchoType] # Classification of each round
|
|
|
|
# Derived metrics
|
|
depth: int = 0 # How many EXPANDS before plateau
|
|
|
|
timestamp: datetime = field(default_factory=datetime.now)
|
|
|
|
|
|
@dataclass
|
|
class ReadinessResult:
|
|
"""Combined analysis for curriculum readiness."""
|
|
|
|
term: str
|
|
level: ReadinessLevel
|
|
action: str # Recommended curriculum action
|
|
|
|
# Supporting evidence
|
|
surface: Optional[SurfaceProbeResult] = None
|
|
echo: Optional[EchoProbeResult] = None
|
|
|
|
# Reasoning
|
|
reasoning: str = ""
|
|
|
|
timestamp: datetime = field(default_factory=datetime.now)
|
|
|
|
def to_dict(self) -> dict:
|
|
"""Convert to JSON-serializable dict."""
|
|
return {
|
|
"term": self.term,
|
|
"readiness": {
|
|
"level": self.level.value,
|
|
"action": self.action,
|
|
"reasoning": self.reasoning,
|
|
},
|
|
"surface": {
|
|
"completions": self.surface.completions if self.surface else [],
|
|
"coherence": self.surface.coherence_score if self.surface else None,
|
|
"hit_eos_count": self.surface.hit_eos_count if self.surface else 0,
|
|
} if self.surface else None,
|
|
"echo": {
|
|
"depth": self.echo.depth if self.echo else 0,
|
|
"types": [t.value for t in self.echo.echo_types] if self.echo else [],
|
|
"chain": self.echo.chain if self.echo else [],
|
|
} if self.echo else None,
|
|
"timestamp": self.timestamp.isoformat(),
|
|
}
|