Skip to content

overview.md

Latest commit

 

History

History
82 lines (59 loc) · 2.51 KB

overview.md

File metadata and controls

82 lines (59 loc) · 2.51 KB
Raw

GeneGIS Architecture Overview

GeneGIS is an AI-native, cloud-native, GPU-native geospatial workbench — not a desktop GIS reimplementation.

Core principle

The center of GeneGIS is Workflow, not Layer.

Traditional:  Data → Layer → Manual Operation → Map
GeneGIS:      Intent → Discovery → Workflow Graph → Verified Execution → Map

Platform layers

Apps (Desktop / Browser / Server / CLI)
  └── UX (Canvas, Command Palette, AI Chat, Workflow Graph)
        └── AI Native Layer (Intent, Planner, Verifier, Provenance)
              └── GIS Kernel
                    ├── Core (project, layers, commands)
                    ├── Spatial / Vector / Raster / PointCloud / Tile
                    ├── Query (DuckDB, PostGIS)
                    ├── Analysis (workflow DAG)
                    └── Rendering (wgpu)
                          └── Data / Storage / Plugin / Collaboration

GIS Core responsibilities

  • Project model, layer graph, CRS metadata, units
  • Command bus (UI, AI, CLI, plugins → same path)
  • Undo/redo, provenance, permissions
  • Workflow graph schema

Implemented in crates/genegis-core and crates/genegis-workflow.

Data philosophy

Rule Meaning
Original data is immutable Sources are referenced, not silently mutated
Derived data is reproducible Outputs trace back to workflow graphs
Cache is disposable Rebuild from workflow + source
Provenance is append-only Audit trail for AI and humans

Cloud-native formats (first-class)

  • Vector: GeoParquet
  • Raster: COG (Cloud Optimized GeoTIFF)
  • Point cloud: COPC
  • Tiles: PMTiles, MBTiles, MVT
  • Catalog: STAC, OGC API

AI native design

AI generates GeoWorkflow IR first — not raw code. Every step carries:

  • Operation name and parameters
  • Expected schema and validation rules
  • CRS / unit requirements
  • Citations and review status

See ai-native.md.

Rendering

wgpu-based pipeline targeting 60fps interactive scenes with:

  • Progressive loading for huge datasets
  • GPU memory budgets per layer
  • LOD for vector, raster tiles, and point clouds

See rendering.md.

Related documents