goon/app/config.py

from functools import lru_cache

from pydantic import Field
from pydantic_settings import BaseSettings, SettingsConfigDict


class Settings(BaseSettings):
    model_config = SettingsConfigDict(env_file=".env", extra="ignore", case_sensitive=False)

    database_url: str = Field(
        default="postgresql+psycopg://goon:goon@localhost:5432/goon",
        validation_alias="DATABASE_URL",
    )

    tpdb_api_token: str | None = Field(default=None, validation_alias="TPDB_API_TOKEN")
    tpdb_base_url: str = Field(
        default="https://api.theporndb.net", validation_alias="TPDB_BASE_URL"
    )

    stashdb_api_key: str | None = Field(default=None, validation_alias="STASHDB_API_KEY")
    stashdb_graphql_url: str = Field(
        default="https://stashdb.org/graphql", validation_alias="STASHDB_GRAPHQL_URL"
    )

    log_level: str = Field(default="INFO", validation_alias="LOG_LEVEL")


    # Sentry observability — pusty DSN = init no-op (devel/local). Cloud free tier
    # 5k errors/mies wystarczy dla 1-user app.
    sentry_dsn: str | None = Field(default=None, validation_alias="SENTRY_DSN")
    sentry_environment: str = Field(default="dev", validation_alias="SENTRY_ENVIRONMENT")
    sentry_traces_sample_rate: float = Field(
        default=0.1, validation_alias="SENTRY_TRACES_SAMPLE_RATE"
    )

    api_keys_raw: str = Field(default="", validation_alias="API_KEYS")
    """Lista API keys oddzielona przecinkami. Pusta = auth wyłączony (tylko dev/local)."""

    allowed_app_sig_hashes_raw: str = Field(default="", validation_alias="ALLOWED_APP_SIG_HASH")
    """Whitelist SHA256 (hex) podpisów APK akceptowane przez backend. Każdy request mobile
    wysyła `X-App-Signature` z hashem signing certu (PackageManager.GET_SIGNING_CERTIFICATES).
    Pusta = check wyłączony (dev/wstępny rollout). Lista = comma-separated lowercase hex.
    Re-packaging APK innym keystorem zmienia hash → 403."""

    auto_merge_threshold: float = 0.92
    review_threshold: float = 0.75
    fingerprint_hamming_max: int = 5
    title_token_set_min: int = 88
    date_window_days: int = 7

    # APScheduler (M5). Każdy 0/None = job wyłączony.
    sched_tpdb_hours: int = Field(default=6, validation_alias="GOON_SCHED_TPDB_HOURS")
    sched_stashdb_hours: int = Field(default=6, validation_alias="GOON_SCHED_STASHDB_HOURS")
    sched_performer_driven_hours: int = Field(
        default=12, validation_alias="GOON_SCHED_PERFORMER_DRIVEN_HOURS"
    )
    sched_performer_driven_top_n: int = Field(
        default=20, validation_alias="GOON_SCHED_PERFORMER_DRIVEN_TOP_N"
    )
    # Continuous worker. interval=15s + max_instances=1 + coalesce=True ⇒ effective rate
    # = max(15, real_tick_duration). Real tick ~50-80s przy full coverage. Set to 0 to disable.
    sched_performer_continuous_seconds: int = Field(
        default=15, validation_alias="GOON_SCHED_PERFORMER_CONTINUOUS_SECONDS"
    )
    sched_performer_continuous_refresh_days: int = Field(
        default=30, validation_alias="GOON_SCHED_PERFORMER_CONTINUOUS_REFRESH_DAYS"
    )
    # Movie ingest — paradisehill (primary) + dooplay mirrory (mangoporn/streamporn/
    # pandamovies). Każdy connector zapisuje swój `Source` i robi delta od ostatniego
    # successful run. Set to 0 to disable. Domyślnie 24h: movie sites rosną wolniej
    # niż tube'y (~5-30 nowych dziennie), nie ma sensu wymiatać częściej.
    sched_movie_ingest_hours: int = Field(
        default=24, validation_alias="GOON_SCHED_MOVIE_INGEST_HOURS"
    )
    # Browse-latest scheduler: freshporno/porn00/pornxp newest scenes.
    # 6h cadence (zmiana z 24h 2026-05-20): user reportował brak Brazzers Exxtra po
    # 15-05. Root cause był 2-fold: (1) freshporno publikuje sceny w ciągu dnia, 24h
    # cadence łapie tylko te do 05:30 UTC; (2) meta_content/release_date bug osobno.
    # 6h = 4 runs/dzień = każda freshporno scena zaingestowana w ciągu ~6h od publik.
    sched_browse_latest_hours: int = Field(
        default=6, validation_alias="GOON_SCHED_BROWSE_LATEST_HOURS"
    )
    sched_browse_latest_max_pages: int = Field(
        default=5, validation_alias="GOON_SCHED_BROWSE_LATEST_MAX_PAGES"
    )
    # Bulk-dedup performers safety net — auto-merge duplikatów które resolver-time
    # scoring pominął. 12h cadence: leci 2x dziennie (po porannym browse-latest run).
    sched_bulk_dedup_hours: int = Field(
        default=12, validation_alias="GOON_SCHED_BULK_DEDUP_HOURS"
    )

    # Hetzner Cloud bandwidth monitor — read-only API token (Security → API Tokens
    # w panelu Hetzner Cloud). Bez tokenu monitor wyłączony (warning w log).
    # Free traffic per server: CX22=20TB, CPX21=20TB itd. Overage = €1/TB.
    hetzner_api_token: str | None = Field(default=None, validation_alias="HETZNER_API_TOKEN")
    hetzner_server_id: int | None = Field(default=None, validation_alias="HETZNER_SERVER_ID")
    # Alert thresholds (% of included_traffic) — Sentry severity levels.
    hetzner_alert_info_pct: int = Field(default=50, validation_alias="HETZNER_ALERT_INFO_PCT")
    hetzner_alert_warning_pct: int = Field(default=80, validation_alias="HETZNER_ALERT_WARNING_PCT")
    hetzner_alert_error_pct: int = Field(default=95, validation_alias="HETZNER_ALERT_ERROR_PCT")

    @property
    def api_keys(self) -> set[str]:
        return {k.strip() for k in self.api_keys_raw.split(",") if k.strip()}

    @property
    def auth_enabled(self) -> bool:
        return bool(self.api_keys)

    @property
    def allowed_app_sig_hashes(self) -> set[str]:
        return {
            h.strip().lower().replace(":", "")
            for h in self.allowed_app_sig_hashes_raw.split(",")
            if h.strip()
        }

    @property
    def app_sig_check_enabled(self) -> bool:
        return bool(self.allowed_app_sig_hashes)


@lru_cache
def get_settings() -> Settings:
    return Settings()