ENTITY test 2 | recorded
Measure the installed ADAM storage matrix
Run the installed ADAM v0.41 audit against repetitive, mixed and unique workloads and preserve exact-view and restart-recovery checks while comparing native factored state with vacuumed SQLite.
Recorded result: 20/20 checks PASS | repetitive 49.311643% lower | mixed 7.620239% lower | unique 75.858036% larger than SQLite
Watch the test
This VP8 WebM evidence replay is generated from the exact published test source and recorded terminal output below.
Run it yourself
python examples/storage_efficiency_test.py
Exact test source
from pathlib import Path
import shutil, sys
INSTALLED = Path(r"E:\ENTITY_ACTIVE\BTG_BUILT_LOCALAPPDATA\BTG\ADAM_RC2\ADAM_v1_0_COMPLETE_SOFTWARE_REFERENCE_RC2")
OUTPUT = Path(r"E:\ENTITY_ACTIVE\ENTITY_TESTING_20261001\runs\storage-efficiency")
sys.path.insert(0, str(INSTALLED))
from adam_v41.audit import run_audit
if OUTPUT.exists():
shutil.rmtree(OUTPUT)
result = run_audit(OUTPUT, records=1500, seed=20260804)
print("ENTITY_STORAGE_TEST=PASS" if result["overall_pass"] else "ENTITY_STORAGE_TEST=FAIL")
print(f"CHECKS={result['checks_passed']}/{result['checks_total']}")
for row in result["workloads"]:
print(f"WORKLOAD={row['mode']}")
print(f"SQLITE_BYTES={row['sqlite_current_bytes']}")
print(f"FACTORED_BYTES={row['adam_native_factored_state_bytes']}")
print(f"FACTORED_REDUCTION_PERCENT={row['native_factored_vs_sqlite_reduction_percent']:.6f}")
print(f"EXACT_VIEW={'PASS' if row['views_equal'] else 'FAIL'}")
print(f"RESTART_RECOVERY={'PASS' if row['checkpoint_restart_pass'] else 'FAIL'}")
Recorded output
ENTITY_STORAGE_TEST=PASS CHECKS=20/20 WORKLOAD=repetitive SQLITE_BYTES=167936 FACTORED_BYTES=85124 FACTORED_REDUCTION_PERCENT=49.311643 EXACT_VIEW=PASS RESTART_RECOVERY=PASS WORKLOAD=mixed SQLITE_BYTES=131072 FACTORED_BYTES=121084 FACTORED_REDUCTION_PERCENT=7.620239 EXACT_VIEW=PASS RESTART_RECOVERY=PASS WORKLOAD=unique SQLITE_BYTES=253952 FACTORED_BYTES=446595 FACTORED_REDUCTION_PERCENT=-75.858036 EXACT_VIEW=PASS RESTART_RECOVERY=PASS
Claim boundary: This test intentionally shows that storage reduction is workload-dependent. The 97.14% Level-3 result does not generalize to SQLite or high-entropy/unique workloads.