Official sample starting states#

Choosing an official sample should open a useful biological view, not make you recreate the same setup before you can explore. Each built-in sample therefore publishes one small, reviewed starting state beside its prepared scientific files. Cellucid applies that state automatically; it is not a second dataset and it is not a manual Load State step.

What happens when you choose a sample#

  1. Cellucid validates and loads the selected prepared dataset.

  2. The positions, fields, metadata, and other scientific data become the live dataset.

  3. Only after the scientific dataset has been published, Cellucid reads the state capability advertised for that exact generation.

  4. It validates the manifest, downloads the one declared state, verifies the state bytes against the catalog’s SHA-256, and confirms that the selected generation has not changed.

  5. Cellucid automatically applies the verified static starting view.

This order matters: an unverified view file is never treated as scientific data. If an advertised manifest, state, or digest is invalid, Cellucid reports that failure after the dataset is available; it does not guess another filename or silently substitute a different preset.

The published default.cellucid-session is an internal five-chunk starting state, not a manual user bundle. Downloading it and choosing Load State is unsupported and the strict generic reader rejects it. To create a portable manual state, choose the sample, wait for this verified view, and then use Save State.

The exact publication contract#

Every official export generation contains both sidecars:

Built-in sample

State bundle

Manifest

Suo

exports/suo/default.cellucid-session

exports/suo/state-snapshots.json

Garcia

exports/garcia/default.cellucid-session

exports/garcia/state-snapshots.json

He

exports/he/default.cellucid-session

exports/he/state-snapshots.json

Kanemaru

exports/kanemaru/default.cellucid-session

exports/kanemaru/state-snapshots.json

Pancreas

exports/pancreas/default.cellucid-session

exports/pancreas/state-snapshots.json

The manifest has one exact root, states, and one exact entry:

{ "states": ["default.cellucid-session"] }

The matching entry in exports/datasets.json advertises state_manifest and state_sha256 as a pair. For example, the Suo entry declares:

{
  "state_manifest": "state-snapshots.json",
  "state_sha256": "5581569118eaa992d36cd3ce53cde0aa488d27b45f89b8f1a41d56de8f7b7a67"
}

state_manifest names the manifest beside that dataset generation. state_sha256 pins the exact bytes of default.cellucid-session; it is not a digest of the JSON manifest. Both fields are required for the capability to be advertised.

You can inspect all five generation directories in the cellucid-datasets exports.

What the reviewed view changes#

Each official state:

  • activates the cell_type observation field;

  • restores a modest, static close-up in the 3-D Orbit camera;

  • uses the light theme and grid background; and

  • starts without active filters or kept snapshots.

These are deliberately static-only presets. They contain no cinematic/camera chunk, so they never restore a recorded Camera Path, Loop playback, or Autoplay. The static camera position can change when the state is applied, but moving-camera playback remains off.

Which catalogs opt in—and which sources are not probed#

Automatic starting states belong to a local-demo catalog generation whose entry explicitly advertises the paired state_manifest and state_sha256 fields. The official catalog uses this capability. A custom exportsBaseUrl catalog may deliberately opt in by publishing the same paired fields and exact sidecars.

Custom hosting is therefore not rejected as a class: only explicit catalog advertisement enables the state requests. Jupyter, GitHub, local-user, and remote sources are not probed, and neither is any catalog entry without the complete capability.

  • Jupyter (jupyter);

  • a GitHub repository (github-repo);

  • a locally selected prepared export (local-user);

  • a remote prepared-data server (remote); and

  • any catalog entry without the complete state capability.

Cellucid does not infer sidecar paths from filenames or look for an unadvertised state. This keeps custom hosting predictable and avoids unrequested network traffic.

For those zero-probe sources, configure the view you want and use the ordinary Save/restore UX (manual save + restore) workflow. Your own .cellucid-session remains a manual, identity-matched working-state file. See Current format and dataset identity for the matching rules.