sheafatlas 0.1.0a1

Typed manifests and deterministic queries for coordinate-aware artifact atlases.

sheafatlas

sheafatlas is a small, dependency-free manifest format for coordinate-aware artifact catalogs. It is designed for systems that need to answer questions such as “which artifacts are valid for this coordinate?” without introducing a database or an ordering convention hidden in application code.

The package is deliberately manifest-first. AtlasItem stores an artifact's stable identifier, human label, coordinates, tags, source, and optional metadata. AtlasManifest provides immutable updates and deterministic lookup. The JSON codec makes manifests easy to commit, review, and exchange between tools.

Quick start

from sheafatlas import AtlasItem, AtlasManifest, dumps, loads

manifest = AtlasManifest(
    "vision-assets",
    (
        AtlasItem(
            "camera/front",
            "Front camera",
            ("image", "rgb"),
            frozenset({"sensor", "primary"}),
            source="capture-lab",
        ),
    ),
    version="2026-01",
)

matches = manifest.search(tags={"sensor"}, coordinate="rgb")
payload = dumps(manifest)
same_manifest = loads(payload)

IDs are unique within a manifest. Searches require every requested tag and return items sorted by ID, so output stays stable across runs. Calling with_item returns a new manifest instead of mutating an existing one.

Manifest shape

The serialized form is ordinary JSON and can be reviewed or generated by other languages:

{
  "items": [
    {
      "coordinates": ["image", "rgb"],
      "id": "camera/front",
      "label": "Front camera",
      "source": "capture-lab",
      "tags": ["primary", "sensor"]
    }
  ],
  "name": "vision-assets",
  "version": "2026-01"
}

CLI

Inspect a manifest file with:

python -m sheafatlas path/to/manifest.json

The library has no runtime dependencies and supports Python 3.10 and newer.