Output

Output#

Configure where model results are published.

Constructor#

Output(
    geocatalog_uri,
    collection_name,
    credential,
    storage_url,
    blob_container,
    storage_account_key=None,
    run_id=None,
    save_local=False,
    output_dir="./output"
)

Required Parameters#

geocatalog_uri : str
GeoCatalog/STAC endpoint URL where results will be published.

collection_name : str
Name of the collection to create or use for results.

credential : Any
Azure credential for GeoCatalog and blob storage authentication.

storage_url : str
Azure Blob Storage account URL (e.g., https://account.blob.core.windows.net).

blob_container : str
Blob container name for storing chip imagery and STAC item assets.

Optional Parameters#

storage_account_key : str | None
Storage account key for SAS generation. Only needed if Azure AD delegation is unavailable.

Recommended: Use Azure RBAC roles instead:

  • Storage Blob Data Contributor (for uploads)

  • Storage Blob Delegator (for SAS generation)

run_id : str | None
Run identifier for organizing results. Auto-generated UUID if not provided.

save_local : bool
Save results to local files for debugging. Default: False.

output_dir : str
Local directory for saved files (only used if save_local=True). Default: ./output.

Publishing Behavior#

GeoCatalog publishing happens automatically when you provide an Output object to model.run().

Results include:

  • STAC items with detection/segmentation metadata

  • Chip imagery and overlays in blob storage

  • GeoJSON features with geometries and properties

Examples#

Standard Configuration#

from azure.identity import DefaultAzureCredential
import geoai

output = geoai.Output(
    geocatalog_uri="https://geocatalog.contoso.com",
    collection_name="building-detections",
    credential=DefaultAzureCredential(),
    storage_url="https://myaccount.blob.core.windows.net",
    blob_container="results"
)

With Local Debugging#

output = geoai.Output(
    geocatalog_uri="https://geocatalog.contoso.com",
    collection_name="road-detections",
    credential=DefaultAzureCredential(),
    storage_url="https://myaccount.blob.core.windows.net",
    blob_container="results",
    save_local=True,
    output_dir="./debug_output"
)

With Storage Account Key (Fallback)#

import os

output = geoai.Output(
    geocatalog_uri="https://geocatalog.contoso.com",
    collection_name="detections",
    credential=DefaultAzureCredential(),
    storage_url="https://myaccount.blob.core.windows.net",
    blob_container="results",
    storage_account_key=os.getenv("STORAGE_ACCOUNT_KEY")  # Fallback auth
)

Blob Storage Layout#

Results are organized in blob storage:

{blob_container}/
└── {model_name}/
    └── {run_id}/
        ├── chips/
        │   ├── chip_0_0.tif
        │   ├── chip_0_1.tif
        │   └── ...
        └── results/
            ├── merged_imagery.tif
            ├── final_overlay.tif
            ├── final_overlay.jpg
            └── detections.geojson

Validation#

Output destinations are validated when you call model.validate_input() or model.run().

Validation checks:

  • GeoCatalog endpoint is reachable

  • Credential has write permissions

  • Blob storage account is accessible

  • Blob container exists or can be created

See Validation for details.

Common Issues#

Error: “Failed to generate SAS token”#

Cause: Missing RBAC roles or storage account key.

Solution: Either:

  1. Assign Azure RBAC roles (recommended):

    • Storage Blob Data Contributor

    • Storage Blob Delegator

  2. Or provide storage_account_key parameter

See Also#

  • Complete Example - Full workflow with Output configuration

  • Input - Configure data sources

  • Validation - Pre-flight checks

Contents