Env vars (NoteTrace-specific)¶

NoteTrace reads the same shared variables as the other Trace apps (PORT, BASE_URL, JWT_SECRET, DB_PATH, UPLOADS_PATH, SMTP_*, OIDC_*, AI_*, INSECURE_COOKIES, LOG_LEVEL, RECOVERY_TOKEN, TOKEN_ENC_KEY, <NAME>_FILE secrets, and the rest). Those live on the shared environment reference. This page lists the values that are specific to NoteTrace or worth calling out for it.

Container¶

Env var	Default	Purpose
`PORT`	`3004`	Port the server listens on inside the container. The sample compose file maps it to the same host port.
`DB_PATH`	`/data/db/notetrace.db`	SQLite database file. Mount `/data/db` as a volume.
`UPLOADS_PATH`	`/data/uploads`	Uploaded images, voice notes, files, and server-side backups. Mount `/data/uploads` as a volume.
`UPLOAD_MAX_MB`	`100`	The largest single upload (an image, a recording, or a file on a note), in MB. A bigger file is refused with a message saying so. Raise your reverse proxy's body limit to match (Nginx `client_max_body_size`).
`BASE_URL`	(empty)	Subpath mount, for example `/notetrace`.

Integrations¶

Env var	Default	Purpose
`WEBHOOKS_ENABLED`	(unset)	Set to `1` to turn on outgoing webhooks (`note.created`, `checklist.completed`, `reminder.fired`).
`ALLOW_PRIVATE_WEBHOOK_URLS`	(unset)	Set to `1` to allow webhook targets on private or loopback addresses, such as a Home Assistant container on the same Docker network.
`MCP_ENABLED` / `MCP_WRITE_ENABLED` / `MCP_DESTROY_ENABLED`	(unset)	Model Context Protocol endpoint and its write and destructive tiers. See MCP.
`ALLOW_PRIVATE_COOKTRACE_URLS`	(unset)	Set to `1` so the CookTrace link can reach a CookTrace on a LAN, loopback, or Docker network address.
`ALLOW_PRIVATE_LINK_PREVIEWS`	(unset)	Set to `1` to show link previews for links to LAN, loopback, or Docker network addresses.

Trace¶

Env var	Default	Purpose
`AI_TRANSCRIBE_MODEL`	provider default	Speech-to-text model for voice notes when Trace is set by env vars: `gpt-4o-mini-transcribe` on OpenAI, `whisper-1` on an OpenAI-compatible server. Gemini uses `AI_MODEL`. Whisper-style models give timestamped transcripts.

Audio¶

The Docker image includes a small audio-only ffmpeg, used to convert voice recordings browsers can't play (Google Keep's 3GP and AMR) and to split long recordings for transcription. Running the server outside Docker, install ffmpeg or point these at it; without it those two features are off and everything else works.

Env var	Default	Purpose
`FFMPEG_PATH`	`ffmpeg`	ffmpeg binary.
`FFPROBE_PATH`	`ffprobe`	ffprobe binary, used to read a recording's length.

Backup¶

Env var	Default	Purpose
`BACKUP_UPLOAD_MAX_MB`	`512`	Upload size cap (MB) for restoring a full-backup zip.
`BACKUP_SCHEDULE`	(unset)	`off`, `daily`, `weekly`, or `monthly`. Locks the Settings field when set.
`BACKUP_TIME`	(unset)	Auto-backup time (`HH:MM`, container time zone). Locks the Settings field.
`BACKUP_RETENTION`	(unset)	How many auto-backups to keep.

Path convenience (Docker Compose only)¶

Env var	Default	Purpose
`DATA_DB_PATH`	(compose)	Host folder for the database volume. Read by the compose file, not by the app.
`DATA_UPLOADS_PATH`	(compose)	Host folder for the uploads volume.

No env vars needed¶

Reminders, sharing, and imports have no server settings. Server reminder delivery runs whenever a user has a push service or a reminder.fired webhook set up. Imports are sent in batches of up to 500 notes, so they fit within the server's normal request limits without tuning.