Custom Datasets

Rasteret’s dataset registry lets you define reusable dataset descriptors for your own STAC APIs, GeoParquet indexes, or local collections. Once registered, you can build Collections using the simple rasteret.build(dataset_id, ...) API.

Quick Start: Register a STAC Dataset

import rasteret
from rasteret import DatasetDescriptor

# Define your dataset
descriptor = DatasetDescriptor(
    id="acme/field-survey-2024",
    name="ACME Field Survey 2024",
    description="High-res drone imagery, 10cm, agricultural fields",
    stac_api="https://stac.acme.com/v1",
    stac_collection="field-survey-2024",
    band_map={
        "R": "red",
        "G": "green",
        "B": "blue",
        "NIR": "nir",
    },
    spatial_coverage="regional",
    temporal_range=("2024-01-01", "2024-12-31"),
    license="proprietary",
)

# Register it
rasteret.register(descriptor)

# Use it
collection = rasteret.build(
    "acme/field-survey-2024",
    name="field-001",
    bbox=(11.3, 48.1, 11.5, 48.3),
    date_range=("2024-06-01", "2024-08-31"),
)

DatasetDescriptor Fields

A DatasetDescriptor captures dataset identity, access methods, and band mappings:

Identity

DatasetDescriptor(
    id="my-org/my-dataset",            # Namespaced ID
    name="My Dataset",                   # Human-readable name
    description="Brief description",     # One-liner
)

Access: STAC API

DatasetDescriptor(
    id="my-org/sentinel-3",
    stac_api="https://stac.example.com/v1",
    stac_collection="sentinel-3-olci",
    band_map={
        "B01": "Oa01",
        "B02": "Oa02",
        # ... more bands
    },
)

Access: GeoParquet

DatasetDescriptor(
    id="my-org/custom-index",
    geoparquet_uri="s3://my-bucket/index.parquet",
    column_map={
        "fid": "id",
        "geom": "geometry",
        "timestamp": "datetime",
    },
    href_column="cog_url",
    band_index_map={"R": 0, "G": 1, "B": 2},  # Multi-band COG
)

Coverage & Licensing

DatasetDescriptor(
    id="my-org/dataset",
    spatial_coverage="global",  # "global", "regional", "local", etc.
    temporal_range=("2020-01-01", "2024-12-31"),
    license="CC-BY-4.0",         # SPDX identifier or "proprietary"
    license_url="https://creativecommons.org/licenses/by/4.0/",
    commercial_use=True,         # False for licenses like CC-BY-NC
)

Authentication

DatasetDescriptor(
    id="my-org/private-data",
    requires_auth=True,
    s3_credentials_url="https://example.com/api/credentials",
    cloud_config={
        "provider": "aws",
        "requester_pays": True,
        "region": "us-west-2",
        "url_patterns": {
            "https://cdn.example.com/": "s3://my-bucket/",
        },
    },
)

Registering Local Collections

If you’ve built a Collection and want to make it available via the registry:

from pathlib import Path

# Build a Collection
collection = rasteret.build_from_stac(
    name="my-local-collection",
    stac_api="...",
    collection="sentinel-2-l2a",
    bbox=(11.3, 48.1, 11.5, 48.3),
    date_range=("2024-01-01", "2024-06-30"),
)

# Export it
export_path = Path("./collections/my_local_collection")
collection.export(export_path)

# Register it
descriptor = rasteret.register_local(
    dataset_id="local/my-collection",
    path=export_path,
    description="My analysis-ready collection",
)

# Now you can load it by ID
reloaded = rasteret.build("local/my-collection", name="reloaded")

What it does:

Creates a DatasetDescriptor pointing to the local Parquet path
Registers it in the global registry (in-memory)
Persists it to ~/.rasteret/datasets.local.json so it’s available in future sessions

Managing the Registry

List All Datasets

from rasteret import DatasetRegistry

for desc in DatasetRegistry.list():
    print(f"{desc.id}: {desc.name}")

Search Datasets

# Search by keyword (case-insensitive)
results = DatasetRegistry.search("sentinel")
for desc in results:
    print(f"{desc.id}: {desc.name}")

Get a Descriptor

desc = DatasetRegistry.get("earthsearch/sentinel-2-l2a")
if desc:
    print(f"Name: {desc.name}")
    print(f"STAC API: {desc.stac_api}")
    print(f"Bands: {list(desc.band_map.keys())}")

Unregister a Dataset

# Unregister from in-memory registry
DatasetRegistry.unregister("my-org/my-dataset")

# Unregister from local persistent registry
from rasteret.catalog import unregister_local_descriptor
unregister_local_descriptor("local/my-collection")

Advanced: Multi-Band COGs

For datasets where multiple bands are stored in a single COG file:

DatasetDescriptor(
    id="my-org/naip-custom",
    stac_api="https://stac.example.com/v1",
    stac_collection="naip-custom",
    band_map={
        "R": "image",   # All bands map to the same asset
        "G": "image",
        "B": "image",
        "NIR": "image",
    },
    band_index_map={  # Sample index within the multi-band COG
        "R": 0,
        "G": 1,
        "B": 2,
        "NIR": 3,
    },
    separate_files=False,  # Indicates multi-band COG
)

Example: Private STAC API

import rasteret
from rasteret import DatasetDescriptor

# Define descriptor
descriptor = DatasetDescriptor(
    id="mycompany/internal-imagery",
    name="Internal High-Res Imagery",
    description="Proprietary 1m imagery, internal use only",
    stac_api="https://stac.internal.mycompany.com/v1",
    stac_collection="internal-2024",
    band_map={
        "R": "red",
        "G": "green",
        "B": "blue",
    },
    spatial_coverage="regional",
    temporal_range=("2024-01-01", "present"),
    requires_auth=True,
    license="proprietary",
    commercial_use=False,
)

# Register it
rasteret.register(descriptor)

# Use with credentials
from obstore.store import S3Store
backend = rasteret.create_backend(
    default_s3_config={"region": "us-east-1"},
)

collection = rasteret.build(
    "mycompany/internal-imagery",
    name="project-x",
    bbox=(-77.1, 38.8, -76.9, 39.0),
    date_range=("2024-06-01", "2024-08-31"),
    backend=backend,
)

Example: GeoParquet Index from Source Cooperative

import rasteret
from rasteret import DatasetDescriptor

# AEF embeddings (already registered, shown here for reference)
descriptor = DatasetDescriptor(
    id="my-org/aef-custom",
    name="AEF Embeddings (Custom)",
    description="64-band int8 embeddings, 10m, annual",
    geoparquet_uri=(
        "s3://us-west-2.opendata.source.coop/"
        "tge-labs/aef/v1/annual/aef_index.parquet"
    ),
    column_map={"fid": "id", "geom": "geometry", "year": "datetime"},
    href_column="path",
    band_index_map={f"A{i:02d}": i for i in range(64)},
    bbox_columns={
        "minx": "wgs84_west",
        "miny": "wgs84_south",
        "maxx": "wgs84_east",
        "maxy": "wgs84_north",
    },
    separate_files=False,
    spatial_coverage="global",
    temporal_range=("2018-01-01", "2023-12-31"),
    license="CC-BY-4.0",
    cloud_config={
        "provider": "aws",
        "region": "us-west-2",
        "url_patterns": {
            "s3://us-west-2.opendata.source.coop/": ("https://data.source.coop/"),
        },
    },
)

rasteret.register(descriptor)

See /home/daytona/workspace/source/examples/aef_duckdb_query.py:1 for a complete example using this descriptor. To share a dataset descriptor with teammates:

Export to JSON

from rasteret.catalog import export_local_descriptor
from pathlib import Path

export_path = export_local_descriptor(
    dataset_id="local/my-collection",
    output_path=Path("./shared/my_collection_descriptor.json"),
)

print(f"Exported to {export_path}")

Import from JSON

import json
from rasteret import DatasetDescriptor

with open("my_collection_descriptor.json") as f:
    data = json.load(f)

descriptor = DatasetDescriptor(**data)
rasteret.register(descriptor)

Best Practices

Namespacing

Use namespaced IDs to avoid collisions:

org/dataset: For organization-owned datasets
provider/collection: For specific STAC provider variants (e.g. pc/sentinel-2-l2a vs earthsearch/sentinel-2-l2a)
local/name: For local collections

Band Naming

Use consistent band codes across datasets:

Sentinel-2: B01, B02, …, B12, SCL
Landsat: B1, B2, …, B7, qa_pixel
Custom: Choose descriptive codes (e.g. R, G, B, NIR)

Documentation

Include clear descriptions and license info:

DatasetDescriptor(
    id="my-org/dataset",
    name="My Dataset",
    description="Brief description (visible in CLI and docs)",
    license="CC-BY-4.0",
    license_url="https://creativecommons.org/licenses/by/4.0/",
)

Example Queries

Provide example bbox/date ranges for testing:

DatasetDescriptor(
    id="my-org/dataset",
    example_bbox=(11.3, 48.1, 11.5, 48.3),
    example_date_range=("2024-06-01", "2024-06-30"),
)

Next Steps

Building Collections - Use registered datasets with rasteret.build()
Cloud Authentication - Set up auth for private datasets
Data Analysis - Analyze pixel data from custom datasets

Get Started

Core Concepts

Guides

Integrations

Advanced

Quick Start: Register a STAC Dataset

DatasetDescriptor Fields

Identity

Access: STAC API

Access: GeoParquet

Coverage & Licensing

Authentication

Registering Local Collections

Managing the Registry

List All Datasets

Search Datasets

Get a Descriptor

Unregister a Dataset

Advanced: Multi-Band COGs

Example: Private STAC API

Example: GeoParquet Index from Source Cooperative

Export to JSON

Import from JSON

Best Practices

Namespacing

Band Naming

Documentation

Example Queries

Next Steps

Build docs developers (and LLMs) love

Get Started

Core Concepts

Guides

Integrations

Advanced

Documentation Index

​Quick Start: Register a STAC Dataset

​DatasetDescriptor Fields

​Identity

​Access: STAC API

​Access: GeoParquet

​Coverage & Licensing

​Authentication

​Registering Local Collections

​Managing the Registry

​List All Datasets

​Search Datasets

​Get a Descriptor

​Unregister a Dataset

​Advanced: Multi-Band COGs

​Example: Private STAC API

​Example: GeoParquet Index from Source Cooperative

​Sharing Descriptors

​Export to JSON

​Import from JSON

​Best Practices

​Namespacing

​Band Naming

​Documentation

​Example Queries

​Next Steps

Build docs developers (and LLMs) love

Quick Start: Register a STAC Dataset

DatasetDescriptor Fields

Identity

Access: STAC API

Access: GeoParquet

Coverage & Licensing

Authentication

Registering Local Collections

Managing the Registry

List All Datasets

Search Datasets

Get a Descriptor

Unregister a Dataset

Advanced: Multi-Band COGs

Example: Private STAC API

Example: GeoParquet Index from Source Cooperative

Sharing Descriptors

Export to JSON

Import from JSON

Best Practices

Namespacing

Band Naming

Documentation

Example Queries

Next Steps