"""Runtime settings (environment variables, `AIA_` prefix). Never log `settings.model_dump()` — it contains secrets."""
from __future__ import annotations

from functools import lru_cache
from pathlib import Path

from pydantic import Field
from pydantic_settings import BaseSettings, SettingsConfigDict


class Settings(BaseSettings):
    model_config = SettingsConfigDict(env_file=(".env",), env_file_encoding="utf-8", extra="ignore")

    app_env: str = Field("development", alias="APP_ENV")
    site_url: str = Field("https://www.ai-atlas.co", alias="AIA_SITE_URL")

    database_url: str = Field("postgresql+asyncpg://aiatlas:aiatlas@127.0.0.1:5432/aiatlas", alias="DATABASE_URL")
    redis_url: str = Field("redis://127.0.0.1:6379/5", alias="REDIS_URL")

    data_dir: Path = Field(Path("./data"), alias="AIA_DATA_DIR")
    api_host: str = Field("127.0.0.1", alias="AIA_API_HOST")
    api_port: int = Field(8321, alias="AIA_API_PORT")
    admin_token: str = Field("", alias="AIA_ADMIN_TOKEN")
    log_json: bool = Field(True, alias="AIA_LOG_JSON")
    tz: str = Field("America/Toronto", alias="AIA_TZ")

    # Crawler
    user_agent: str = Field("AIAtlasBot/0.1 (+https://www.ai-atlas.co/bot; contact@spboucher.ai)", alias="AIA_USER_AGENT")
    http_timeout_s: float = Field(45.0, alias="AIA_HTTP_TIMEOUT")
    max_body_bytes: int = Field(25 * 1024 * 1024, alias="AIA_MAX_BODY_BYTES")
    default_rate_per_min: int = Field(30, alias="AIA_DEFAULT_RATE_PER_MIN")
    respect_robots: bool = Field(True, alias="AIA_RESPECT_ROBOTS")
    fetch_concurrency: int = Field(8, alias="AIA_FETCH_CONCURRENCY")

    # Escalation transports (optional; never required)
    scrapfly_api_key: str = Field("", alias="SCRAPFLY_API_KEY")
    firecrawl_api_key: str = Field("", alias="FIRECRAWL_API_KEY")
    browser_enabled: bool = Field(False, alias="AIA_BROWSER_ENABLED")

    # Local LLM factory (OpenAI-compatible endpoint, e.g. MacLustr llm-api.io). Optional.
    llm_base_url: str = Field("", alias="AIA_LLM_BASE_URL")
    llm_api_key: str = Field("", alias="AIA_LLM_API_KEY")
    llm_small_model: str = Field("qwen3-4b-instruct-2507-4bit", alias="AIA_LLM_SMALL_MODEL")
    llm_medium_model: str = Field("qwen3.6-35b-a3b-4bit", alias="AIA_LLM_MEDIUM_MODEL")
    llm_large_model: str = Field("qwen3.8-27b-4bit", alias="AIA_LLM_LARGE_MODEL")
    llm_timeout_s: float = Field(600.0, alias="AIA_LLM_TIMEOUT")
    llm_enabled: bool = Field(True, alias="AIA_LLM_ENABLED")
    embedding_model: str = Field("qwen3-embedding-0.6b-8bit", alias="AIA_EMBEDDING_MODEL")
    embedding_dim: int = Field(1024, alias="AIA_EMBEDDING_DIM")

    # Scheduler
    scheduler_tick_s: int = Field(30, alias="AIA_SCHEDULER_TICK_S")
    worker_concurrency: int = Field(4, alias="AIA_WORKER_CONCURRENCY")
    backup_cron: str = Field("40 4 * * *", alias="AIA_BACKUP_CRON")

    @property
    def raw_dir(self) -> Path:
        return self.data_dir / "raw"

    @property
    def text_dir(self) -> Path:
        return self.data_dir / "text"

    @property
    def logs_dir(self) -> Path:
        return self.data_dir / "logs"

    @property
    def backups_dir(self) -> Path:
        return self.data_dir / "backups"

    @property
    def cache_dir(self) -> Path:
        return self.data_dir / "cache"

    @property
    def sync_database_url(self) -> str:
        return self.database_url.replace("+asyncpg", "")

    @property
    def llm_available(self) -> bool:
        return bool(self.llm_enabled and self.llm_base_url and self.llm_api_key)

    def ensure_dirs(self) -> None:
        for d in (self.raw_dir, self.text_dir, self.logs_dir, self.backups_dir, self.cache_dir, self.data_dir / "seed"):
            d.mkdir(parents=True, exist_ok=True)


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


settings = get_settings()