internet_use_control/windows/webrestrict/settings.py

"""Client settings, read from one JSON file (AGENTS.md §11 Phase 4).

Everything the client needs is local configuration: the broker address, the
device's own topic and MQTT account, and where to keep its state. Nothing here
is a second source of truth — the Control Service still owns the rules.

The file is provisioned on the machine and is **never** committed. It holds the
device's MQTT password, which must stay off the repository (AGENTS.md §10 rule
8). The path comes from ``--config``, then ``WEBRESTRICT_CONFIG``, then
``%PROGRAMDATA%\\WebRestrict\\client.json``.
"""

from __future__ import annotations

import json
import os
from dataclasses import dataclass, field
from pathlib import Path
from typing import Any

# Default location. ProgramData is machine-wide and readable by the New Comp
# account, which is what the logon Scheduled Task runs as.
DEFAULT_CONFIG_NAME = "client.json"
DEFAULT_STATE_DIR_NAME = "WebRestrict"

# Never kill these, whatever a blocklist says. The client must not be able to
# take the machine down, and it must not kill itself (it runs as python.exe).
# The client's own interpreter and its parent chain are added at runtime.
DEFAULT_PROTECTED: tuple[str, ...] = (
    "system",
    "system idle process",
    "registry",
    "smss",
    "csrss",
    "wininit",
    "winlogon",
    "services",
    "lsass",
    "svchost",
    "dwm",
    "explorer",
    "taskhostw",
    "sihost",
    "ctfmon",
    "conhost",
    "fontdrvhost",
    "audiodg",
    "msmpeng",
    "securityhealthservice",
    "searchindexer",
    "startmenuexperiencehost",
    "shellexperiencehost",
    "textinputhost",
    "applicationframehost",
    "python",
    "pythonw",
)


def default_program_data() -> Path:
    """The machine-wide directory for client configuration and state."""
    base = os.environ.get("PROGRAMDATA") or os.environ.get("ALLUSERSPROFILE")
    if base:
        return Path(base) / DEFAULT_STATE_DIR_NAME
    # Not Windows (unit tests, a lab box): keep it under the user's home.
    return Path.home() / ".webrestrict"


def default_config_path() -> Path:
    """The config path: ``WEBRESTRICT_CONFIG`` wins, else ``%PROGRAMDATA%``."""
    override = os.environ.get("WEBRESTRICT_CONFIG")
    if override:
        return Path(override)
    return default_program_data() / DEFAULT_CONFIG_NAME


@dataclass
class Settings:
    """Everything the client needs to run, with safe lab defaults.

    A missing field falls back to the default, so a hand-written config file
    only needs the values that differ. Unknown fields are ignored, so the file
    may grow without a client update.
    """

    # --- the link ---------------------------------------------------------
    broker_host: str = "192.168.20.13"
    broker_port: int = 8883
    ca_file: str = ""  # empty = the platform trust store
    device_topic: str = ""
    mqtt_username: str = ""
    mqtt_password: str = ""
    client_id: str = ""  # empty = derive from the topic token

    # --- local state ------------------------------------------------------
    state_dir: str = ""

    # --- timings (seconds) ------------------------------------------------
    poll_seconds: float = 5.0
    telemetry_seconds: float = 60.0
    status_seconds: float = 60.0
    connect_timeout_seconds: float = 10.0

    # --- behaviour --------------------------------------------------------
    overlay: bool = True
    protected: list[str] = field(default_factory=lambda: list(DEFAULT_PROTECTED))

    # ---------------------------------------------------------------------
    @property
    def state_path(self) -> Path:
        """The directory holding the cached config, the usage roll-up and the log."""
        if self.state_dir:
            return Path(self.state_dir)
        if self.device_token:
            return default_program_data() / self.device_token
        return default_program_data()

    @property
    def device_token(self) -> str:
        """The middle topic level, which is also the broker account name."""
        base = self.device_topic.removesuffix("/config")
        return base.rsplit("/", 1)[-1] if "/" in base else base

    @property
    def status_topic(self) -> str:
        return self.device_topic.removesuffix("/config") + "/status"

    @property
    def telemetry_topic(self) -> str:
        return self.device_topic.removesuffix("/config") + "/telemetry"

    @property
    def effective_client_id(self) -> str:
        return self.client_id or f"kid-{self.device_token or 'windows'}"

    def config_path(self) -> Path:
        return self.state_path / "config-cache.json"

    def telemetry_path(self) -> Path:
        return self.state_path / "telemetry.json"

    def log_path(self) -> Path:
        return self.state_path / "client.log"

    def validate(self) -> None:
        """Raise when the client cannot possibly work, instead of failing later."""
        if not self.device_topic:
            raise ValueError("device_topic is not set (e.g. restrict/finn_s_computer_2/config)")
        if not self.device_topic.endswith("/config"):
            raise ValueError(f"device_topic must end with /config: {self.device_topic!r}")
        if not self.broker_host:
            raise ValueError("broker_host is not set")
        if not 1 <= self.broker_port <= 65535:
            raise ValueError(f"broker_port is out of range: {self.broker_port}")
        if not self.mqtt_username:
            raise ValueError("mqtt_username is not set (it is the topic token)")

    # ---------------------------------------------------------------------
    @classmethod
    def from_dict(cls, raw: dict[str, Any]) -> "Settings":
        """Build settings from a parsed JSON object, ignoring unknown keys."""
        known = {f for f in cls.__dataclass_fields__}  # type: ignore[attr-defined]
        data = {k: v for k, v in raw.items() if k in known}
        settings = cls(**data)
        settings.broker_port = int(settings.broker_port)
        return settings

    @classmethod
    def load(cls, path: str | os.PathLike[str] | None = None) -> "Settings":
        """Read the config file and validate it."""
        target = Path(path) if path else default_config_path()
        with open(target, "r", encoding="utf-8") as handle:
            raw = json.load(handle)
        if not isinstance(raw, dict):
            raise ValueError(f"{target}: expected a JSON object")
        settings = cls.from_dict(raw)
        settings.validate()
        return settings