# ---- API ----------------------------------------------------------------
# Port the Hono API listens on.
PORT=3000

# Database driver: sqlite (default) | postgres | mysql
DB_DRIVER=sqlite

# SQLite: path to the database file (created on first migrate).
DATABASE_PATH=./data/bookguardian.db

# Postgres / MySQL: connection URL (only read when DB_DRIVER matches).
# DATABASE_URL=postgres://user:password@localhost:5432/bookguardian
# DATABASE_URL=mysql://user:password@localhost:3306/bookguardian

# Timezone the API uses for "today" — the default read date when a book is
# marked as read and the "no future dates" check. Match your phone's timezone,
# especially in Docker, where the container defaults to UTC.
# TZ=Europe/Madrid

# Serve the built web app (apps/web/dist) from the API as well, so one process
# is one origin. Leave unset for `pnpm dev`; the Docker image sets /app/web.
# WEB_DIST=../web/dist

# ---- Web build ----------------------------------------------------------
# Build-time only (read by apps/web/vite.config.ts, never at runtime): the
# origin baked into the landing page's canonical, Open Graph and Twitter card
# tags, which crawlers want absolute because they do not run the SPA. There is
# no default host: unset means relative tags (/ and /og.png), which are correct
# everywhere but give a chat unfurl no absolute image URL. Every real
# deployment should set it to the same value as AUTH_BASE_URL. docker compose
# picks it up from here as a build arg; plain Docker: --build-arg PUBLIC_ORIGIN=...
# PUBLIC_ORIGIN=https://books.example.com

# ---- Accounts (Google sign-in) ------------------------------------------
# OAuth client of type "Web application" from Google Cloud Console → APIs &
# Services → Credentials (see README "Google sign-in"). Without these two the
# API boots but /api/auth/google answers 503 auth_not_configured.
# GOOGLE_CLIENT_ID=
# GOOGLE_CLIENT_SECRET=
# Public origin people open the app at (no trailing slash). The redirect URI
# to register in Google Cloud is exactly ${AUTH_BASE_URL}/api/auth/google/callback.
# Cookies are marked Secure when this is https. Default: http://localhost:$PORT;
# for `pnpm dev` use the Vite origin (http://localhost:5173), which proxies /api.
# AUTH_BASE_URL=https://books.example.com
# Signs the short-lived cookie that carries the in-flight sign-in state
# (≥ 32 characters, e.g. `openssl rand -base64 32`). Generated per process
# when unset, which only means a restart interrupts sign-ins in progress.
# AUTH_COOKIE_SECRET=
# Optional allow-list: only these emails may create an account (existing
# accounts keep working). Comma-separated (spaces, semicolons or newlines are
# accepted too), case-insensitive. Unset = anyone with a Google account can
# sign in — set it on a personal instance.
# AUTH_ALLOWED_EMAILS=you@example.com,partner@example.com
# Session lifetime in days (sliding: renewed on use past the halfway point).
# AUTH_SESSION_DAYS=30

# ---- Web ----------------------------------------------------------------
# Leave unset: the Vite dev server and `vite preview` proxy /api to the API,
# so the SPA uses relative URLs and its session cookie travels with them.
# Setting VITE_API_URL bakes an absolute API origin into the bundle, which is
# only right for a static deployment against a remote API on the same site.
# VITE_API_URL=http://localhost:3000
# Dev/preview proxy target when the API is not on http://localhost:3000.
# API_PROXY_TARGET=http://localhost:3000

# ---- Book metadata lookup -----------------------------------------------
# Open Library is queried first, Google Books when it has no record.
# OPEN_LIBRARY_URL=https://openlibrary.org
# GOOGLE_BOOKS_URL=https://www.googleapis.com/books/v1
# Optional. Google's anonymous quota is tiny; create a key in Google Cloud
# (Books API). With a key, Google Books is also the second source for covers
# (better coverage for Spanish editions); without one that step is skipped.
# Read by the API process: `.env` here or in apps/api/, or a runtime variable
# in Docker / Coolify. Never baked into the image.
# GOOGLE_BOOKS_API_KEY=
# Per-provider request budget and how long free-text search results stay
# cached in memory. ISBN lookups are not cached in memory: they are persisted
# in the shared `catalog_books` table (see below).
# LOOKUP_TIMEOUT_MS=8000
# LOOKUP_CACHE_TTL_SECONDS=86400

# ---- Shared ISBN catalogue ----------------------------------------------
# Metadata fetched for an ISBN is stored in the database, shared by every user
# and survives restarts, so each ISBN is fetched online once. Rows older than
# CATALOG_REFRESH_DAYS are served immediately and refreshed in the background;
# an ISBN no provider knows is retried after CATALOG_MISS_DAYS.
# CATALOG_REFRESH_DAYS=180
# CATALOG_MISS_DAYS=7

# ---- Book covers --------------------------------------------------------
# Covers are WebP files on disk (never rows in the database), one per image
# content hash, shared by every book with the same ISBN. Directory defaults
# to `covers` next to the SQLite file (/data/covers in Docker).
# COVERS_DIR=./data/covers
# Image host of Open Library (the metadata host is OPEN_LIBRARY_URL above).
# OPEN_LIBRARY_COVERS_URL=https://covers.openlibrary.org
# How long "no provider has a cover for this ISBN" is remembered.
# COVERS_MISS_DAYS=30
# Unreferenced shared covers are deleted once older than this (daily GC).
# COVERS_GC_DAYS=90
# Pause between provider requests; Open Library asks for at most 1 per second.
# COVERS_MIN_INTERVAL_MS=1000