"""Stage a verified public release using an explicit allowlist, without credentials."""

from __future__ import annotations
import argparse
import hashlib
import json
import re
import shutil
import xml.etree.ElementTree as ET
from datetime import datetime, timezone
from pathlib import Path
from codelock_affect.vectors import VectorBank

ROOT_FILES = (
    "README.md",
    "LICENSE",
    "NOTICE",
    "pyproject.toml",
    "requirements.txt",
    "Makefile",
    "MANIFEST.in",
    ".gitignore",
    ".gitattributes",
    "adapter-release.json",
    "adapter_config.json",
    "adapter_model.safetensors",
    "added_tokens.json",
    "chat_template.jinja",
    "merges.txt",
    "special_tokens_map.json",
    "tokenizer.json",
    "tokenizer_config.json",
    "vocab.json",
)
PUBLIC_DIRS = (
    "codelock_affect",
    "scripts",
    "tests",
    "configs",
    "data",
    "docs",
    "vectors",
    "results",
)
SUFFIXES = {".py", ".json", ".jsonl", ".md", ".txt", ".sh", ".safetensors", ".xml"}


def release_files(root):
    paths = [root / name for name in ROOT_FILES if (root / name).is_file()]
    for directory in PUBLIC_DIRS:
        paths.extend(
            path
            for path in (root / directory).rglob("*")
            if path.is_file()
            and path.suffix in SUFFIXES
            and not any(
                part.startswith(".") or part == "__pycache__"
                for part in path.relative_to(root).parts
            )
        )
    return sorted(paths, key=lambda path: path.relative_to(root).as_posix())


def build_release(root, output=None):
    root = Path(root).resolve()
    bank = VectorBank.load(root / "vectors/multichannel")
    if set(bank.vectors) != {"pain", "fear", "anxiety", "relief"}:
        raise ValueError("The v0.2 release requires all four channels")
    data_hash = hashlib.sha256(
        (root / "data/affect_pairs.jsonl").read_bytes()
    ).hexdigest()
    if any(
        vector.meta.get("dataset_file_sha256") != data_hash
        for vector in bank.vectors.values()
    ):
        raise ValueError("Calibration vectors do not match the current dataset")
    suites = ET.parse(root / "results/v2/tests.xml").getroot().iter("testsuite")
    counts = [dict(suite.attrib) for suite in suites]
    if not counts or any(
        int(suite.get("failures", 0)) or int(suite.get("errors", 0)) for suite in counts
    ):
        raise ValueError("A passing test report is required")
    replay = json.loads((root / "results/v2/replay.json").read_text(encoding="utf-8"))
    if not replay.get("complete") or any(
        "generation_context" not in row for row in replay.get("steps", [])
    ):
        raise ValueError("The event-grounded replay is incomplete")
    evaluation = json.loads(
        (root / "results/v2/evaluation.json").read_text(encoding="utf-8")
    )
    if not evaluation.get("complete"):
        raise ValueError("Calibration evaluation must be complete")
    for seed in (7, 23, 41):
        result = json.loads(
            (root / f"results/v2/recommended-seed{seed}.json").read_text(
                encoding="utf-8"
            )
        )
        if not result.get("complete"):
            raise ValueError(f"Recommended evaluation seed {seed} is incomplete")
    stamp = datetime.now(timezone.utc).strftime("%Y%m%d-%H%M%S")
    output = (
        Path(output).resolve()
        if output
        else root / ".local" / "release-staging" / stamp
    )
    if output.exists() and any(output.iterdir()):
        raise ValueError(
            "Release output must be empty; an earlier bundle will not be overwritten"
        )
    # Keep generated files outside the public allowlist to avoid recursive staging.
    if output.is_relative_to(root) and output.relative_to(root).parts[0] in PUBLIC_DIRS:
        raise ValueError("Release staging cannot be inside a public source directory")
    output.mkdir(parents=True, exist_ok=True)
    manifest = {
        "version": "0.2.0",
        "created_utc": datetime.now(timezone.utc).isoformat(),
        "base_model": "Qwen/Qwen3-8B",
        "contains_base_weights": False,
        "files": {},
    }
    token_pattern = re.compile(
        r"(?:hf_[A-Za-z0-9]{20,}|tsk_[A-Za-z0-9_-]{20,}|-----BEGIN (?:OPENSSH |RSA |EC )?PRIVATE KEY-----)"
    )
    for path in release_files(root):
        if path.is_symlink() or not path.resolve().is_relative_to(root):
            raise ValueError(
                f"Public file must remain inside the source directory: {path.name}"
            )
        payload = path.read_bytes()
        if path.suffix != ".safetensors" and token_pattern.search(
            payload.decode("utf-8", errors="replace")
        ):
            raise ValueError(
                f"Credential-like content found in {path.relative_to(root)}; refusing publication"
            )
        relative = path.relative_to(root)
        target = output / relative
        target.parent.mkdir(parents=True, exist_ok=True)
        shutil.copy2(path, target)
        manifest["files"][relative.as_posix()] = {
            "sha256": hashlib.sha256(payload).hexdigest(),
            "bytes": len(payload),
        }
    (output / "release-manifest.json").write_bytes(
        (json.dumps(manifest, indent=2) + "\n").encode("utf-8")
    )
    return output, manifest


def main():
    p = argparse.ArgumentParser(description=__doc__)
    p.add_argument("--output")
    args = p.parse_args()
    root = Path(__file__).resolve().parents[1]
    output, manifest = build_release(root, args.output)
    print(
        json.dumps(
            {
                "staging": str(output),
                "files": len(manifest["files"]),
                "bytes": sum(row["bytes"] for row in manifest["files"].values()),
            },
            indent=2,
        )
    )


if __name__ == "__main__":
    main()