Skip to content

ROADMAP.md

Latest commit

 

History

History
48 lines (36 loc) · 2.32 KB

ROADMAP.md

File metadata and controls

48 lines (36 loc) · 2.32 KB
Raw

Roadmap

This roadmap focuses on making Limen more useful, reproducible and welcoming as an open-source landslide-risk monitoring framework.

Near term

  • Improve onboarding docs for local Docker, Neon/dev and self-hosted deployment.
  • Add more good first issue tasks for docs, tests, data adapters and frontend map improvements.
  • Document each public data source with license, cadence, spatial resolution and ingestion assumptions.
  • Add reproducible examples for one small AOI and one regional AOI.
  • Expand tests around scoring thresholds, regional configuration and alert deduplication.

OSS readiness

  • Keep contribution, security, issue and PR templates up to date.
  • Document architecture invariants and scoring assumptions in reviewer-friendly language.
  • Track external contributions and credit contributors in release notes.
  • Keep labels consistent: good first issue, help wanted, documentation, geospatial, backend, frontend, data-source, scoring, security.

Scientific and operational reliability

  • Publish validation notebooks/reports for reproducible backtests.
  • Make uncertainty and limitations visible in API responses and UI copy.
  • Add more explainability examples for individual cells and AOIs.
  • Separate official-source data from derived/modelled indicators in the UI and docs.
  • Strengthen monitoring around stale data, ingestion failures and drift.

Reusable components

Potential reusable packages or extractable modules:

  • geospatial ingestion helpers for Italian public hazard datasets;
  • interpretable multi-factor scoring primitives;
  • PostGIS grid and vector-tile utilities;
  • alert deduplication and notification adapters;
  • MCP tools for geospatial risk summaries.

Long term

  • Dynamic/forecast flood component (post-V2): combine forecast rainfall with the static ISPRA hazard class into a flood_forecast factor feeding the reserved hydrology weight — deterministic first, pure scoring, opt-in feed. Design recorded in issue #8; 2D hydraulic modelling stays out of scope.
  • More regional validation cases.
  • Cleaner package boundaries for data ingestion and scoring.
  • Public demo datasets suitable for contributors without large downloads.
  • Contributor-friendly benchmarks for speed, accuracy and false-alarm tradeoffs.
  • Documentation for municipalities, researchers and civic-tech groups.