Skip to content

configuration.md

Latest commit

 

History

History
45 lines (36 loc) · 2.29 KB

configuration.md

File metadata and controls

45 lines (36 loc) · 2.29 KB
Raw

Configuration

All configuration is through environment variables, read from .env at startup (via dotenv) or from the process environment. Copy .env.example to .env and fill in the values.

Migrations run automatically at startup, so a reachable DATABASE_URL is brought up to date without a manual step.

Environment variables

"Required" means the process — or the relevant feature — panics or refuses to operate without the variable. Variables scoped to a feature only matter when that feature is compiled in.

Var Scope Required? Notes
BOT_TOKEN core yes Discord bot token. Secret.
DATABASE_URL core yes Postgres DSN. Migrations auto-run.
DB_USER / DB_PASSWORD / DB_NAME / DB_NETWORK / DB_PORT core compose Composed into DATABASE_URL for Docker Compose.
RUST_LOG core no Overrides the default tracing filter.
AI_MODEL ai yes (if AI built) Panics if unset when AI is compiled in.
AI_API_KEY ai (hosted) yes for hosted backends Not needed for ai-ollama. Secret.
AI_MAX_MSG_CONTEXT ai no (default 10, cap 100) Context window size.
AI_CHAT_ENDPOINT ai-ollama only no Override the Ollama base URL.
REDIS_URL redis / ai no Enables cross-instance context, locks, and limits.
TOTAL_SHARDS / SHARD_START / SHARD_END sharding no All three or none. Inclusive range.

Secrets

BOT_TOKEN and AI_API_KEY are secrets. None of these are ever logged. See SECURITY.md for the full inventory and handling.

Feature-specific notes

  • ai is a meta-feature: enabling it also enables redis. It does not pick a provider — you must enable exactly one ai-<backend>. See the README feature matrix and docs/ai.md.
  • util-download needs yt-dlp and ffmpeg (with ffprobe) on PATH at runtime. Optionally set YT_DLP_COOKIES_PATH to a Netscape-format cookies file to authenticate with YouTube and bypass bot detection.
  • tokio_console additionally needs RUSTFLAGS="--cfg tokio_unstable" at build time. See docs/observability.md.
  • Sharding is opt-in: set all three of TOTAL_SHARDS, SHARD_START, and SHARD_END, or none. See docs/deployment.md.