slate.tracker / docs

Configuration Reference

The full set of settings in config.yaml. Anything not listed here is reserved for future use; setting it has no effect.

File location

Windows: %APPDATA%\SlateTracker\config.yaml
macOS: ~/Library/Application Support/SlateTracker/config.yaml
Linux (community build): ~/.config/slate-tracker/config.yaml

The agent watches the file and reloads on save. Invalid values fall back to the previous valid value and surface an error in the local dashboard.

Top level structure

license:
  key: "<your license key>"

watch:
  folder: "C:/Users/<you>/Pictures/references"
  recursive: true
  ignore_dotfiles: true

output:
  root: "D:/References/Episode 04"
  folder_pattern: "{scene}/{lens}"
  filename_pattern: "{scene}_{take}_{timestamp}.jpg"
  on_collision: "suffix"
  constants:
    ep: "04"

vision:
  model: standard
  max_parallel: 3
  resize_long_edge: 1600

review:
  min_confidence: 80
  required_fields: ["scene", "take"]

desk:
  console: "grandma3"
  host: "10.0.5.20"
  port: 30000
  user_variable: "SlateTracker.LatestReference"

dashboard:
  port: 7745
  bind: "127.0.0.1"

telemetry:
  crash_reports: true
  usage_metrics: false

license

Key	Type	Default	Description
`key`	string	required	Your license key. Without it the agent runs in offline preview only.

watch

Key	Type	Default	Description
`folder`	string	required	The folder to watch for new images.
`recursive`	boolean	`true`	Watch subfolders.
`ignore_dotfiles`	boolean	`true`	Skip files whose name starts with a dot.

output

Key	Type	Default	Description
`root`	string	required	The folder all processed files end up under.
`folder_pattern`	string	`"{scene}/{lens}"`	Subfolder pattern relative to `root`. See Folder Structure.
`filename_pattern`	string	`"{scene}_{take}_{timestamp}.{ext}"`	Filename pattern.
`on_collision`	enum	`"suffix"`	One of `suffix`, `overwrite`, `skip`.
`constants`	map	`{}`	Extra placeholders, available as `{name}` in patterns.

vision

Key	Type	Default	Description
`model`	string	`"standard"`	Friendly tier name. `standard` is the default; `high` is the stronger model for tough slates. You can also pass an exact provider model string if you know what you want.
`max_parallel`	integer	`3`	How many images to process at once. Higher values are faster but spike CPU.
`resize_long_edge`	integer	`1600`	Long edge in pixels for the image sent to the recognition engine. Lower is cheaper, less accurate.

review

Key	Type	Default	Description
`min_confidence`	integer	`80`	Below this confidence the entry goes to the review queue.
`required_fields`	string[]	`["scene","take"]`	Entries missing any of these fields land in review even at high confidence.

desk

Key	Type	Default	Description
`console`	enum	`"none"`	One of `none`, `grandma3`, `grandma2` (beta), `etc_eos` (beta).
`host`	string	`""`	IP or hostname of the console.
`port`	integer	`30000`	TCP port. Defaults match the chosen console.
`user_variable`	string	`"SlateTracker.LatestReference"`	grandMA user variable name.
`cue_label_pattern`	string	`"{scene}_{take}"`	Pattern used to match incoming images to existing cues.

dashboard

Key	Type	Default	Description
`port`	integer	`7745`	Port the local dashboard listens on.
`bind`	string	`"127.0.0.1"`	Address to bind the local dashboard to. Set to `0.0.0.0` to expose on the LAN.

telemetry

Key	Type	Default	Description
`crash_reports`	boolean	`true`	Send crash backtraces to help us fix bugs.
`usage_metrics`	boolean	`false`	Send anonymised counts of processed images.

Validation

When you save, the agent runs a quick validation:

Required fields are present.
Folder paths exist and are writable.
Pattern placeholders are known.
Network settings parse.

If anything fails, the agent reverts to the previous config and shows the error on the local dashboard with a line number.