Skip to content

CHART documentation

CHART uses one Python API, Dagster for background data work, Keycloak for access, Postgres/PostGIS for saved data and results, and the canonical Next web app for the browser experience.

This site explains how to run the platform locally and operate its application, model, and climate data pipeline.

CHART is developed as open-source infrastructure under the GNU Affero General Public License v3.0. The project is designed for inspection, local adaptation, and independent deployment while supporting digital-public-good principles.

Start here

Follow Getting started to install CHART, start the complete local environment, and verify that it is working.

Local and published addresses

This documentation site is public. Addresses beginning with 127.0.0.1 refer to services on your own computer and only work after you start CHART locally.

Runtime

CHART has one Python core (core/chart/):

  • FastAPI for synchronous app and engine endpoints
  • Dagster for batch climate ingestion and model handoffs
  • SQLAlchemy + Alembic as the sole schema owner

The retired Fastify/Drizzle API is not installed, started, or deployed.

How the pieces connect

Dagster does not own application requests or user-facing API routes. FastAPI saves durable work in Postgres, and Dagster claims that work, prepares the required climate inputs, and hands validated inputs to the model. The public action repository remains a separate publishing service that CHART reads over HTTP.

flowchart LR
    repository["Public action repository<br/>API or snapshot"] --> adapter["FastAPI repository adapter"]
    adapter --> web["CHART web"]

    web --> api["FastAPI"]
    api --> requests[("Postgres<br/>saved requests")]
    requests --> dagster["Dagster sensor and job"]
    providers["Climate data providers"] --> dagster
    dagster --> inputs[("Postgres<br/>validated climate inputs")]
    inputs --> model["Versioned model"]
    model --> results[("Postgres<br/>saved results")]
    results --> api
Page Use it for
Getting started Install and run CHART locally
Installation setup Complete the first-run onboarding wizard, reset an installation, or recover from a stuck bootstrap
Sign-in and email How people sign in with emailed codes, and the SMTP settings (SES, Resend, Mailpit) that deliver them
Users and capabilities (design) Agreed design: workspaces, roles as capability bundles, editable departments, invitations and one authorisation check
Python core overview Build FastAPI modules
Modeling Understand the LBW model, inputs, outputs, and limitations
Climate API Understand preview and prediction parameters
API explorer Browse the published API contracts
Data pipeline Run Dagster assets and materialisation
Add a geography and model Extend the supported analytical areas
Where model updates happen Register, version, or replace a health model — for CHART devs and external adopters
Digital public good and licensing Understand CHART's AGPL licence, source availability, and third-party boundaries