Environment Variables¶

Complete reference for all cast2md configuration options.

All settings use uppercase environment variable names. They can be set in a .env file or as system environment variables.

Storage¶

Variable	Default	Description
`STORAGE_PATH`	`./data/podcasts`	Base directory for audio and transcripts
`TEMP_DOWNLOAD_PATH`	`./data/temp`	Temporary directory for downloads in progress

Variable	Default	Description
`DATABASE_URL`	(required)	PostgreSQL connection string
`POSTGRES_PASSWORD`	(required for Docker)	Password for Docker Compose PostgreSQL

Variable	Default	Description
`WHISPER_MODEL`	`large-v3-turbo`	Model identifier (see Whisper Models)
`WHISPER_DEVICE`	`auto`	Device: `cpu`, `cuda`, or `auto`
`WHISPER_COMPUTE_TYPE`	`int8`	Precision: `int8`, `float16`, `float32`
`WHISPER_BACKEND`	`auto`	Backend: `auto`, `faster-whisper`, `mlx`
`WHISPER_CHUNK_THRESHOLD_MINUTES`	`30`	Episodes longer than this are chunked
`WHISPER_CHUNK_SIZE_MINUTES`	`30`	Size of each chunk
`TRANSCRIPTION_BACKEND`	`whisper`	Primary backend: `whisper` or `parakeet`

Variable	Default	Description
`MAX_CONCURRENT_DOWNLOADS`	`2`	Maximum simultaneous audio downloads
`MAX_TRANSCRIPT_DOWNLOAD_WORKERS`	`4`	Parallel workers for external transcripts
`MAX_RETRY_ATTEMPTS`	`3`	Maximum retries for failed downloads
`REQUEST_TIMEOUT`	`30`	HTTP request timeout in seconds

Variable	Default	Description
`STUCK_THRESHOLD_MINUTES`	`30`	Jobs running longer than this are marked stuck

Variable	Default	Description
`TRANSCRIPT_UNAVAILABLE_AGE_DAYS`	`14`	Episodes older than this without external transcript URLs are marked `needs_audio`
`TRANSCRIPT_RETRY_DAYS`	`14`	How long to retry external transcript downloads

Variable	Default	Description
`ITUNES_COUNTRY`	`de`	ISO 3166-1 alpha-2 country code for iTunes lookups

Variable	Default	Description
`NTFY_ENABLED`	`false`	Enable ntfy.sh notifications
`NTFY_URL`	`https://ntfy.sh`	ntfy service URL
`NTFY_TOPIC`	(empty)	ntfy topic name (required if enabled)

Variable	Default	Description
`DISTRIBUTED_TRANSCRIPTION_ENABLED`	`false`	Enable remote transcription nodes
`NODE_HEARTBEAT_TIMEOUT_SECONDS`	`60`	Seconds before marking a node offline
`REMOTE_JOB_TIMEOUT_MINUTES`	`30`	Maximum time for remote transcription jobs

These are never stored in the database:

Variable	Default	Description
`RUNPOD_API_KEY`	(empty)	RunPod API key
`RUNPOD_TS_AUTH_KEY`	(empty)	Tailscale auth key for pod networking

Variable	Default	Description
`RUNPOD_ENABLED`	`false`	Enable RunPod GPU workers
`RUNPOD_MAX_PODS`	`3`	Maximum concurrent GPU pods
`RUNPOD_AUTO_SCALE`	`false`	Auto-start pods on queue growth
`RUNPOD_SCALE_THRESHOLD`	`10`	Queue depth to trigger auto-scaling
`RUNPOD_PODS_PER_THRESHOLD`	`1`	Pods to start per threshold crossed
`RUNPOD_IDLE_TIMEOUT_MINUTES`	`10`	Auto-terminate idle pods (0 to disable)
`RUNPOD_GPU_TYPE`	`NVIDIA RTX A5000`	Preferred GPU type
`RUNPOD_BLOCKED_GPUS`	`NVIDIA GeForce RTX 4090,...`	Comma-separated GPU blocklist
`RUNPOD_WHISPER_MODEL`	`parakeet-tdt-0.6b-v3`	Default model for pods
`RUNPOD_IMAGE_NAME`	`meltforce/cast2md-afterburner:cuda124`	Docker image for pods
`RUNPOD_TS_HOSTNAME`	`runpod-afterburner`	Base Tailscale hostname
`RUNPOD_GITHUB_REPO`	`meltforce/cast2md`	GitHub repo for pod code updates
`RUNPOD_SERVER_URL`	(empty)	Server URL for pod registration
`RUNPOD_SERVER_IP`	(empty)	Tailscale IP of server

These apply to transcriber node workers (including RunPod pods):

Variable	Default	Description
`NODE_REQUIRED_EMPTY_CHECKS`	`2`	Empty queue checks before termination
`NODE_EMPTY_QUEUE_WAIT`	`60`	Seconds between empty queue checks
`NODE_IDLE_TIMEOUT_MINUTES`	`10`	Terminate after no jobs processed (0 to disable)
`NODE_SERVER_UNREACHABLE_MINUTES`	`5`	Terminate after server unreachable
`NODE_MAX_CONSECUTIVE_FAILURES`	`3`	Terminate after consecutive failures (0 to disable)
`NODE_PERSISTENT`	`0`	Set to `1` to disable all auto-termination

Variable	Default	Description
`TS_AUTHKEY`	(empty)	Tailscale auth key
`TAILSCALE_HOSTNAME`	`cast2md`	Tailscale hostname