mirror of
https://github.com/allaunthefox/Research-Stack.git
synced 2026-07-31 03:05:21 +00:00
92 lines
3.2 KiB
Python
92 lines
3.2 KiB
Python
#!/usr/bin/env python3
|
|
"""
|
|
Global Sovereign Manifold Audit
|
|
RGFlow on Unified Compression Data
|
|
"""
|
|
|
|
import sys
|
|
import pandas as pd
|
|
import numpy as np
|
|
from pathlib import Path
|
|
import logging
|
|
|
|
# Add parent directory to path
|
|
sys.path.insert(0, str(Path(__file__).parent.parent.parent))
|
|
sys.path.insert(0, str(Path(__file__).parent.parent.parent / "4-Infrastructure"))
|
|
sys.path.insert(0, str(Path(__file__).parent.parent.parent / "0-Core-Formalism"))
|
|
|
|
from scripts.rgflow_blind_detector import BlindDetector
|
|
|
|
logging.basicConfig(level=logging.ERROR)
|
|
|
|
def run_global_audit(input_parquet: Path, output_summary: Path):
|
|
print(f"Opening Master Sovereign Manifold: {input_parquet.name}")
|
|
df = pd.read_parquet(input_parquet)
|
|
|
|
detector = BlindDetector()
|
|
results = []
|
|
|
|
print(f"Auditing {len(df)} records across Unified Compression stack...")
|
|
|
|
# We iterate through the records. We need to find the 'data' or 'ids' column.
|
|
data_col = next((c for c in df.columns if c in ['input_ids', 'ids', 'data', 'text']), df.columns[0])
|
|
|
|
# Audit a representative sample if too large
|
|
sample_size = min(len(df), 20000)
|
|
df_sample = df.sample(sample_size, random_state=42)
|
|
|
|
for i, (idx, row) in enumerate(df_sample.iterrows()):
|
|
content = row['content']
|
|
seq = str(content)
|
|
|
|
# 1. Mutation (mu): high character transitions in code
|
|
transitions = sum(1 for k in range(len(seq)-1) if seq[k] != seq[k+1])
|
|
mu_q = (transitions / len(seq)) * 0.1 if len(seq) > 0 else 0
|
|
|
|
# 2. Entropy
|
|
counts = np.unique(list(seq), return_counts=True)[1]
|
|
probs = counts / len(seq)
|
|
entropy = -np.sum(probs * np.log2(probs + 1e-9))
|
|
|
|
# 3. Admissibility
|
|
# Mathlib code is dense. Sigma should be high.
|
|
sigma_q = 1.0 + (entropy / 4.0)
|
|
|
|
state = detector.calculate_window_state('ACGT'*100)
|
|
state.sigma_q = sigma_q
|
|
state.mu_q = mu_q
|
|
|
|
(lawful_now, lawful_under_flow, _, _, _, _, _, rg_depth, attractor_id, _) = \
|
|
detector.adaptation_eq.evaluate_state(state)
|
|
|
|
if i < 5:
|
|
print(f"Record {idx}: mu={mu_q:.4f}, sigma={sigma_q:.4f}, lawful={lawful_under_flow}")
|
|
|
|
results.append({
|
|
"idx": idx,
|
|
"lawful": lawful_under_flow,
|
|
"sigma": sigma_q,
|
|
"depth": rg_depth,
|
|
"path": row.get('path', 'unknown')
|
|
})
|
|
|
|
if i % 1000 == 0:
|
|
print(f"Audit Progress: {i}/{sample_size}")
|
|
|
|
res_df = pd.DataFrame(results)
|
|
top_lawful = res_df[res_df['lawful'] == True].sort_values(by='sigma', ascending=False)
|
|
|
|
print("\n--- TOP LAWFUL SIGNALS IN UNIFIED STACK ---")
|
|
print(top_lawful.head(10))
|
|
|
|
top_lawful.to_csv(output_summary, index=False)
|
|
print(f"\nAudit complete. Summary saved to {output_summary}")
|
|
|
|
if __name__ == "__main__":
|
|
master_p = Path("/home/allaun/Documents/Research Stack/data/datasets/master_sovereign_manifold.parquet")
|
|
output_s = Path("/home/allaun/Documents/Research Stack/data/datasets/global_audit_summary.csv")
|
|
|
|
if master_p.exists():
|
|
run_global_audit(master_p, output_s)
|
|
else:
|
|
print("Error: Master Manifold not found.")
|