Mutable state shared across UI components.

Batch inference over DuckDB embeddings. Scores all embeddings in batches using a trained classifier and returns IDs where probability ex

Loads training data from GeoJSON and fetches embeddings from DuckDB. Supports two matching modes: 1. tile_id matching: GeoJSON has tile_

Manages the tile-based results pane.

Generates GeoJSON output from classification detections.

Timing information for batch inference.

XGBoost binary classifier for embeddings.

Timing information for data loading operations.

Timing information for output generation.

Evaluation metrics for binary classification.

Exception raised for errors in the ingestion of Parquet files.

Comprehensive timing for entire pipeline.

Results from classification pipeline run.

Cross-validation results with mean and std for each metric.

Example MGRS tile: 18SJH

Configuration for negative sampling.

Orchestrates the full classification workflow. Usage: with ClassificationPipeline(duckdb_path="path/to/db.duckdb") as pipeline:

Configuration for ESRI LULC data source.

Centralizes status/operation messages displayed in the UI.

Encapsulates configuration, database, and FAISS operations.

Minimal configuration for GeoVibes.

Class for tracking a MGRS tile grid

Responsible for the ipyleaflet map, layers, and status widgets.

Sample negative training examples using Earth Engine stratified sampling. Uses ESRI Global LULC to sample points stratified by landcover cla

Basemap configuration and tile URLs.

Database-related constants.

Interactive map interface for geospatial similarity search.

Map layer styling constants.

Complete configuration for negative sampling pipeline.

Tests for BatchInference class.

Integration tests for BatchInference with real DuckDB.

Tests for combine_datasets function.

Edge case tests for combine_datasets function.

Tests for data loader helper functionality.

Integration tests for ClassificationDataLoader with real DuckDB.

Tests for detection mode in tile panel.

Tests for EmbeddingClassifier.

Tests for OutputGenerator class.

Integration tests for OutputGenerator with real DuckDB.

Integration tests for end-to-end classification pipeline.

Tests for PipelineResult dataclass.

Tests for PipelineTiming dataclass.

Tests for random negative sampling from DuckDB.

Tests for spatial cross-validation.

UI-related constants.