scripts/validate_local_package.py

from __future__ import annotations

import json
import re
from collections import Counter
from pathlib import Path

import pandas as pd

ROOT = Path(__file__).resolve().parents[1]
BANNED_NAMES = {".env", ".DS_Store", "__pycache__"}
BANNED_PATTERNS = [
    r"sk-or-v1-[A-Za-z0-9._-]+",
    r"ghp_[A-Za-z0-9]{20,}",
    r"hf_[A-Za-z0-9]{20,}",
    r"AKIA[0-9A-Z]{16}",
    r"Bearer [A-Za-z0-9._=-]{30,}",
    r"/Users/",
    r"results/validation_120",
    r"results/scaling_507",
    r"data/benchmark/full_507_prompt_packets.jsonl",
    r"data/benchmark/validation_120_prompt_packets.jsonl",
    r"raw_outputs",
]

REQUIRED = [
    "README.md",
    "LICENSE",
    "CITATION.cff",
    "benchmark_data_card.md",
    "FILE_MANIFEST.md",
    "checksums.sha256",
    "schemas/input_packet_schema.json",
    "schemas/output_schema.json",
    "schemas/label_schema.json",
    "data/stableeval_120_enriched/case_list.csv",
    "data/stableeval_120_enriched/labels.csv",
    "data/stableeval_120_enriched/prompt_packets.jsonl",
    "data/stableeval_507_natural/case_list.csv",
    "data/stableeval_507_natural/labels.csv",
    "data/stableeval_507_natural/prompt_packets.jsonl",
    "data/stableeval_507_natural/full_case_pool.csv",
    "results/stableeval_507_natural/final_tables/stableeval_507_natural_v6_leaderboard.csv",
]

SYSTEMS_120 = ["nanobot", "langgraph", "hermes_ai", "evoagentx", "openclaw", "nanoclaw"]
SYSTEMS_507 = [
    "nanobot_qwen_507",
    "langgraph_qwen_507",
    "hermes_ai_qwen_507",
    "evoagentx_qwen_507",
    "openclaw_adapter_qwen_507",
    "nanoclaw_adapter_qwen_507",
]


def count_jsonl(path: Path) -> int:
    count = 0
    with path.open("r", encoding="utf-8") as handle:
        for line_no, line in enumerate(handle, start=1):
            if line.strip():
                json.loads(line)
                count += 1
    return count


def assert_file(path: str) -> Path:
    p = ROOT / path
    if not p.exists():
        raise SystemExit(f"Missing required file: {path}")
    return p


for required in REQUIRED:
    assert_file(required)

for path in ROOT.rglob("*"):
    if path.name in BANNED_NAMES or path.suffix == ".pyc":
        raise SystemExit(f"Banned cache/private file present: {path.relative_to(ROOT)}")
    if path.is_file() and path.stat().st_size > 75 * 1024 * 1024:
        raise SystemExit(f"Unexpected very large file: {path.relative_to(ROOT)}")
    if path.is_file() and path.suffix.lower() in {".md", ".txt", ".json", ".jsonl", ".csv", ".cff", ".sha256"}:
        text = path.read_text(encoding="utf-8", errors="ignore")
        for pattern in BANNED_PATTERNS:
            if re.search(pattern, text):
                raise SystemExit(f"Banned/stale/private pattern in {path.relative_to(ROOT)}: {pattern}")

labels_120 = pd.read_csv(assert_file("data/stableeval_120_enriched/labels.csv"))
labels_507 = pd.read_csv(assert_file("data/stableeval_507_natural/labels.csv"))
cases_120 = pd.read_csv(assert_file("data/stableeval_120_enriched/case_list.csv"))
cases_507 = pd.read_csv(assert_file("data/stableeval_507_natural/case_list.csv"))
leaderboard = pd.read_csv(assert_file("results/stableeval_507_natural/final_tables/stableeval_507_natural_v6_leaderboard.csv"))

if len(cases_120) != 120 or len(labels_120) != 120:
    raise SystemExit("StableEval-120-Enriched count check failed.")
if len(cases_507) != 507 or len(labels_507) != 507:
    raise SystemExit("StableEval-507-Natural count check failed.")
if Counter(labels_507["enhanced_stability_label"]) != Counter({"stable": 441, "watch": 45, "stress": 21}):
    raise SystemExit("507 label distribution check failed.")
if count_jsonl(assert_file("data/stableeval_120_enriched/prompt_packets.jsonl")) != 120:
    raise SystemExit("120 prompt packet count check failed.")
if count_jsonl(assert_file("data/stableeval_507_natural/prompt_packets.jsonl")) != 507:
    raise SystemExit("507 prompt packet count check failed.")
if len(leaderboard) != 16:
    raise SystemExit("507 v6 leaderboard row count check failed.")

for system in SYSTEMS_120:
    if count_jsonl(assert_file(f"results/stableeval_120_enriched/raw_predictions/{system}.jsonl")) != 120:
        raise SystemExit(f"120 raw prediction count failed for {system}")
    if count_jsonl(assert_file(f"results/stableeval_120_enriched/calibrated_predictions/{system}.jsonl")) != 120:
        raise SystemExit(f"120 calibrated prediction count failed for {system}")

for system in SYSTEMS_507:
    if count_jsonl(assert_file(f"results/stableeval_507_natural/raw_predictions/{system}.jsonl")) != 507:
        raise SystemExit(f"507 raw prediction count failed for {system}")
    if count_jsonl(assert_file(f"results/stableeval_507_natural/calibrated_predictions/{system}.jsonl")) != 507:
        raise SystemExit(f"507 calibrated prediction count failed for {system}")

for schema in ["input_packet_schema.json", "output_schema.json", "label_schema.json"]:
    json.loads((ROOT / "schemas" / schema).read_text(encoding="utf-8"))
json.loads((ROOT / "croissant.json").read_text(encoding="utf-8"))

print("Local StableEval Arena Hugging Face package validation passed.")
print(f"120 cases: {len(cases_120)}")
print(f"507 cases: {len(cases_507)}")
print(f"507 label distribution: {dict(Counter(labels_507['enhanced_stability_label']))}")
print(f"507 leaderboard rows: {len(leaderboard)}")