Product vision

A connected workspace for spatial intelligence.

Sprouti is being designed as a modular platform that connects data management, AI assistance, geospatial processing, code execution and application publishing.

◫

Spatial workspace

Interactive project maps, layer grouping, styling, filtering, legends, collaborative context and reusable project templates.

✦

AI spatial assistant

Natural-language exploration, metadata-aware query planning, explainable tool selection and approval-led execution.

◇

Model Builder

Visual workflows that combine datasets, parameters, SQL, built-in tools and custom Python functions.

{ }

Python sandboxes

Isolated browser-based coding and asynchronous jobs with resource limits, dependency controls, logs and reusable outputs.

⌗

Spatial data services

PostGIS-backed vector analysis, cloud object storage, raster tiling, metadata intelligence and API access.

▦

Application delivery

Publish maps, analytical dashboards, APIs and purpose-built geospatial experiences for wider audiences.

Architecture principles

Built to remain open and extensible.

  • MapLibre and open web mapping standards
  • PostGIS as a spatial source of truth
  • Python-first analytical extensibility
  • API-driven modular services
  • Inspectable AI plans and outputs
  • Usage-aware, asynchronous execution
  • Cloud, private-cloud and local deployment options
  • Customer-controlled data and retention policies
Responsible AI

Human review remains part of the workflow.

Sprouti’s intended workflow separates planning from execution. Users can inspect selected data, tools, parameters and generated SQL or code before approving computational work.

Capabilities shown describe the product direction and active development scope. Availability will vary during private pilots.
Security by architecture

Private GeoAI is a deployment model, not a marketing label.

Sprouti is intended to let each organisation choose where data is stored, where models execute and which services are permitted to receive information.

Data residency and isolation

Project-level isolation, organisation-controlled storage locations and deployment in selected regions or local infrastructure.

Model choice and routing

Use approved hosted models, private model endpoints or locally deployed open models. Administrators can control which models and tools are available.

Minimal data exposure

Metadata-aware planning can avoid sending full datasets to an AI model. Only the information required for an approved task should be exposed.

Identity and permissions

Role-based access control, least-privilege service permissions, project membership rules and integration with organisational identity providers.

Protected execution

Sandboxed jobs, network restrictions, resource limits, approved dependencies, secret isolation and separation between planning and execution.

Traceability and retention

Versioned workflows, audit events, provenance and configurable retention. Customer content is not used to train shared models.

Security capabilities will be validated during private pilots. Exact controls will depend on deployment mode, hosting configuration and customer requirements.