urg-forecast-core 0.1.0__tar.gz

urg_forecast_core-0.1.0/.gitignore

@@ -0,0 +1,40 @@
+# Python
+__pycache__/
+*.py[cod]
+*.egg-info/
+.pytest_cache/
+.ruff_cache/
+.mypy_cache/
+.venv/
+venv/
+
+# Build
+dist/
+build/
+
+# Data and models (reference repo: no real data committed)
+/data/raw/
+/data/processed/
+/data/external/
+/models/
+/outputs/
+
+# Parquet files by default are not committed, except the synthetic fixtures
+# (the test fixture and the compact demo fixture shipped inside the package).
+*.parquet
+!tests/fixtures/*.parquet
+!src/urgencias_core/data/_fixtures/*.parquet
+
+# Editor / OS
+.DS_Store
+.idea/
+.vscode/
+*.swp
+
+# Agent/harness scratch (git worktrees created by tooling)
+.claude/worktrees/
+
+# Env
+.env
+.env.*
+!.env.example

urg_forecast_core-0.1.0/CHANGELOG.md

@@ -0,0 +1,48 @@
+# Changelog
+
+All notable changes to this project are documented here. The format is based on
+[Keep a Changelog](https://keepachangelog.com/en/1.1.0/), and this project
+adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html) (API may
+change between minor versions while on 0.x).
+
+## [Unreleased]
+
+## [0.1.0] - 2026-07-20
+
+First public release — the reference pipeline packaged for `pip install`.
+
+### Added
+
+- Core library: visit-level → hourly occupancy time series (event-cumsum),
+  Chilean calendar features, empirical LOS sampler, Monte Carlo occupancy
+  simulation, `Forecaster` protocol, seasonal-naive / statsforecast / LightGBM
+  quantile forecasters, and an evaluation harness.
+- DEIS MINSAL *Atenciones de Urgencia* fetcher with an offline snapshot
+  fallback, bundled inside the package.
+- Minimal FastAPI reference server with server-rendered charts.
+- Console entry points: `urgencias-demo-synthetic`, `urgencias-demo-deis`,
+  `urgencias-server`.
+- Optional-dependency extras: `models`, `viz`, `server`, `fetch`, `all`. The
+  core install ships only pandas/numpy/pyarrow/holidays/pydantic; gated modules
+  raise an actionable error naming the extra to install.
+- `py.typed` marker, PyPI classifiers/keywords/URLs, and packaged demo
+  datasets so the server and demos run out of the box from an installed wheel.
+- CI: lint gate, Python 3.11/3.12 test matrix, build + `twine check`, and a
+  core-only install job.
+
+### Fixed
+
+- Default server fixture resolved a repo-relative path (`parents[3]`) that
+  escaped `site-packages` in an installed wheel; it now resolves the bundled
+  fixture via `importlib.resources`.
+- DEIS weekly aggregation (`demos.deis._to_weekly`) now trims partial weeks at
+  both series edges, not only a fully-empty trailing week. A most-recent week
+  with a missing day (near-real-time reporting lag) was kept and appeared as a
+  spurious end-of-series drop that contaminated the backtest.
+
+### Changed
+
+- Unified formatting on `ruff format` (dropped black).
+
+[Unreleased]: https://github.com/nicoveraz/urg-forecast-core/compare/v0.1.0...HEAD
+[0.1.0]: https://github.com/nicoveraz/urg-forecast-core/releases/tag/v0.1.0

urg_forecast_core-0.1.0/CITATION.cff

@@ -0,0 +1,32 @@
+cff-version: 1.2.0
+message: "If you use this software, please cite it using these metadata."
+title: "urg-forecast-core"
+abstract: >-
+  Open reference pipeline for Chilean emergency department analytics,
+  probabilistic forecasting, and Monte Carlo occupancy simulation: a
+  visits-to-hourly-occupancy transform, Chilean calendar features, a uniform
+  quantile Forecaster interface (seasonal-naive, statsforecast, LightGBM), an
+  evaluation harness, an empirical-LOS Monte Carlo engine, a DEIS MINSAL
+  open-data client, and a minimal reference dashboard.
+type: software
+authors:
+  - given-names: "Nicolás"
+    family-names: "Vera Zúñiga"
+    orcid: "https://orcid.org/0009-0007-9249-3736"
+    affiliation: "Eunosia, Frutillar, Chile"
+version: "0.1.0"
+date-released: "2026-07-20"
+license: MIT
+repository-code: "https://github.com/nicoveraz/urg-forecast-core"
+url: "https://pypi.org/project/urg-forecast-core/"
+keywords:
+  - emergency department
+  - emergency medicine
+  - forecasting
+  - time series
+  - Monte Carlo simulation
+  - capacity planning
+  - Chile
+  - DEIS
+# After the first Zenodo release, add the concept DOI here, e.g.:
+# doi: "10.5281/zenodo.XXXXXXX"

urg_forecast_core-0.1.0/CONTRIBUTING.md

@@ -0,0 +1,72 @@
+# Contributing
+
+`urg-forecast-core` is an open foundation maintained on a best-effort basis. Fork
+freely, adapt it to your hospital, ship it inside your own product. Pull requests
+and issues are welcome; while on 0.x the API may change between minor versions.
+If you need commercial support or bespoke work on top of this foundation, contact
+the author.
+
+## Development
+
+```bash
+uv sync --all-extras --dev   # full toolchain + all optional deps
+uv run pytest -q             # tests
+uv run ruff check .          # lint
+uv run ruff format .         # format (ruff is the single formatter)
+pre-commit install           # optional: run lint/format on commit
+```
+
+The package uses a src layout. Heavy dependencies live behind extras
+(`models`, `viz`, `server`, `fetch`, `all`); modules that need them guard the
+import and raise an actionable error. Keep that pattern when adding code that
+depends on an optional library.
+
+## Cutting a release
+
+Releases publish to PyPI via [Trusted Publishing](https://docs.pypi.org/trusted-publishers/)
+(OIDC) from `.github/workflows/publish.yml` — no API tokens as secrets.
+
+One-time setup: on PyPI, add this repository and the `pypi` environment as a
+trusted publisher for the `urg-forecast-core` project. Optionally validate metadata
+first by publishing a pre-release (e.g. `0.1.0rc1`) to TestPyPI.
+
+Per release:
+
+1. Bump `version` in `pyproject.toml` (semver).
+2. Move the `## [Unreleased]` notes in `CHANGELOG.md` under a new
+   `## [X.Y.Z] - YYYY-MM-DD` heading and update the compare links.
+3. Commit, then tag and push:
+
+   ```bash
+   git tag vX.Y.Z
+   git push origin main --tags
+   ```
+
+The `publish` workflow then runs tests, verifies the tag matches the package
+version, builds, publishes to PyPI, and creates a GitHub Release with the
+changelog section as its notes.
+
+## Archiving on Zenodo (citable DOI)
+
+Releases are archived on [Zenodo](https://zenodo.org/) for a citable DOI.
+Metadata for the archive comes from `.zenodo.json`; `CITATION.cff` provides the
+citation shown on GitHub.
+
+One-time setup:
+
+1. Sign in to Zenodo with GitHub and, under *GitHub* settings, flip the switch
+   **on** for the `urg-forecast-core` repository.
+2. Cut a release (the tag flow above creates a GitHub Release). Zenodo detects
+   the published GitHub Release, archives the source, and mints two DOIs: a
+   **concept DOI** (always resolves to the latest version) and a
+   **version DOI** (this specific release).
+
+After the first release:
+
+3. Add the concept DOI to `CITATION.cff` (`doi:` field), the README "Citing"
+   section (DOI badge), and `paper/paper.md` if submitting to JOSS.
+
+For submitting the software paper (JOSS), the archived Zenodo DOI is required at
+submission; draft papers are under `paper/`. Author ORCID and affiliation are
+set in `paper/paper.md`, `CITATION.cff`, and `.zenodo.json`; keep them in sync
+if authorship changes.

urg_forecast_core-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Nicolás Vera Z.
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

urg_forecast_core-0.1.0/PKG-INFO

@@ -0,0 +1,287 @@
+Metadata-Version: 2.4
+Name: urg-forecast-core
+Version: 0.1.0
+Summary: Reference code for Chilean ED analytics, simulation, and forecasting. Open foundation of Eunosia.
+Project-URL: Homepage, https://github.com/nicoveraz/urg-forecast-core
+Project-URL: Repository, https://github.com/nicoveraz/urg-forecast-core
+Project-URL: Issues, https://github.com/nicoveraz/urg-forecast-core/issues
+Author: Nicolás Vera Z.
+License: MIT License
+        
+        Copyright (c) 2026 Nicolás Vera Z.
+        
+        Permission is hereby granted, free of charge, to any person obtaining a copy
+        of this software and associated documentation files (the "Software"), to deal
+        in the Software without restriction, including without limitation the rights
+        to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+        copies of the Software, and to permit persons to whom the Software is
+        furnished to do so, subject to the following conditions:
+        
+        The above copyright notice and this permission notice shall be included in all
+        copies or substantial portions of the Software.
+        
+        THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+        IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+        FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+        AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+        LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+        OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+        SOFTWARE.
+License-File: LICENSE
+Keywords: chile,deis,emergency-department,forecasting,healthcare,monte-carlo,simulation,time-series
+Classifier: Development Status :: 4 - Beta
+Classifier: Intended Audience :: Healthcare Industry
+Classifier: Intended Audience :: Science/Research
+Classifier: License :: OSI Approved :: MIT License
+Classifier: Operating System :: OS Independent
+Classifier: Programming Language :: Python :: 3
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Topic :: Scientific/Engineering
+Classifier: Topic :: Scientific/Engineering :: Information Analysis
+Classifier: Typing :: Typed
+Requires-Python: <3.13,>=3.11
+Requires-Dist: holidays>=0.50
+Requires-Dist: numpy>=1.26
+Requires-Dist: pandas>=2.2
+Requires-Dist: pyarrow>=16.0
+Requires-Dist: pydantic>=2.6
+Provides-Extra: all
+Requires-Dist: fastapi>=0.110; extra == 'all'
+Requires-Dist: httpx>=0.27; extra == 'all'
+Requires-Dist: jinja2>=3.1; extra == 'all'
+Requires-Dist: lightgbm>=4.3; extra == 'all'
+Requires-Dist: matplotlib>=3.8; extra == 'all'
+Requires-Dist: scikit-learn>=1.4; extra == 'all'
+Requires-Dist: statsforecast>=1.7; extra == 'all'
+Requires-Dist: tabulate>=0.9; extra == 'all'
+Requires-Dist: uvicorn>=0.29; extra == 'all'
+Provides-Extra: fetch
+Requires-Dist: httpx>=0.27; extra == 'fetch'
+Provides-Extra: models
+Requires-Dist: lightgbm>=4.3; extra == 'models'
+Requires-Dist: scikit-learn>=1.4; extra == 'models'
+Requires-Dist: statsforecast>=1.7; extra == 'models'
+Provides-Extra: server
+Requires-Dist: fastapi>=0.110; extra == 'server'
+Requires-Dist: jinja2>=3.1; extra == 'server'
+Requires-Dist: matplotlib>=3.8; extra == 'server'
+Requires-Dist: tabulate>=0.9; extra == 'server'
+Requires-Dist: uvicorn>=0.29; extra == 'server'
+Provides-Extra: viz
+Requires-Dist: matplotlib>=3.8; extra == 'viz'
+Requires-Dist: tabulate>=0.9; extra == 'viz'
+Description-Content-Type: text/markdown
+
+# urg-forecast-core
+
+[![ci](https://github.com/nicoveraz/urg-forecast-core/actions/workflows/ci.yml/badge.svg)](https://github.com/nicoveraz/urg-forecast-core/actions/workflows/ci.yml)
+[![PyPI](https://img.shields.io/pypi/v/urg-forecast-core.svg)](https://pypi.org/project/urg-forecast-core/)
+[![Python](https://img.shields.io/pypi/pyversions/urg-forecast-core.svg)](https://pypi.org/project/urg-forecast-core/)
+[![License: MIT](https://img.shields.io/badge/License-MIT-yellow.svg)](https://github.com/nicoveraz/urg-forecast-core/blob/main/LICENSE)
+
+Reference code for Chilean emergency department (ED) analytics, simulation, and
+forecasting. Open foundation of **Eunosia**.
+
+*Leer en [español](https://github.com/nicoveraz/urg-forecast-core/blob/main/README.es.md).*
+
+It turns visit-level ED data into hourly occupancy series, adds Chilean calendar
+features, forecasts arrivals/occupancy with quantile bands, and runs a Monte
+Carlo census simulator — plus a reader for the public DEIS MINSAL dataset and a
+minimal web dashboard.
+
+## Install
+
+```bash
+pip install urg-forecast-core            # core library (data, timeseries, features, simulation)
+pip install "urg-forecast-core[all]"     # everything: models, viz, server, and data fetchers
+```
+
+The core install is intentionally light (pandas/numpy/pyarrow/holidays/pydantic).
+Heavier capabilities live behind extras — a module that needs one raises a clear
+"install the extra" error:
+
+| Extra | Pulls in | Enables |
+|---|---|---|
+| `models` | lightgbm, statsforecast, scikit-learn | LightGBM/statsforecast quantile forecasters |
+| `viz` | matplotlib, tabulate | charts and markdown tables (demos, server) |
+| `server` | fastapi, uvicorn, jinja2 (+ `viz`) | the reference dashboard server |
+| `fetch` | httpx | DEIS MINSAL and Open-Meteo network clients |
+| `all` | all of the above | the demos and the full test suite |
+
+## Quickstart
+
+After `pip install "urg-forecast-core[all]"`, three console commands are available.
+They run on datasets bundled with the package and write to `./outputs`:
+
+```bash
+urgencias-demo-synthetic          # full pipeline incl. simulation → 3 PNGs
+urgencias-demo-deis --offline     # forecasting on real DEIS hospital data
+urgencias-server                  # reference dashboard at http://127.0.0.1:8000
+```
+
+From a source checkout with [uv](https://docs.astral.sh/uv/):
+
+```bash
+git clone https://github.com/nicoveraz/urg-forecast-core
+cd urg-forecast-core
+uv sync --all-extras
+
+uv run urgencias-demo-synthetic
+uv run urgencias-demo-deis --offline
+uv run urgencias-server
+```
+
+The synthetic demo runs in seconds on a compact one-year synthetic fixture
+(~13,000 visits) bundled with the package. The DEIS demo fetches real public
+data (or falls back to the bundled offline snapshot) and writes backtesting
+tables plus 6-month-ahead forecasts.
+
+## What the pipeline produces
+
+The figures below are the direct output of running the two demos against the
+bundled data.
+
+### 1. From visits to an hourly occupancy series
+
+`urgencias_core.data.timeseries` converts the visit-level parquet (one row per
+visit, with arrival and discharge timestamps) into an hourly census series using
+the event-cumsum trick (+1 at arrival, −1 at discharge, cumulative sum reindexed
+to the hourly grid). The figure shows the last week of the synthetic fixture: a
+clear diurnal cycle with overnight troughs and evening peaks.
+
+![Synthetic hourly occupancy](https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/demo_synthetic_occupancy.png)
+
+### 2. Forecasting layer — 48-hour hourly forecast
+
+On that series, `SeasonalNaiveBaseline` produces a 48-hour hourly forecast with
+P50–P80 and P80–P95 quantile bands. The stronger models (`AutoARIMA`,
+`AutoETS`, `LGBQuantile`) are compared in the evaluation harness and live behind
+the same `Forecaster` interface.
+
+![Synthetic 48h hourly forecast](https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/demo_synthetic_forecast.png)
+
+### 3. Monte Carlo simulation engine
+
+`urgencias_core.simulation.engine` takes future arrivals (sampled from the
+forecast) and, for each arrival, samples an empirical length-of-stay conditional
+on (acuity, arrival hour). Iterating M replicates yields census uncertainty
+bands 24 hours ahead — the basis for surge decisions and shift tables.
+
+![24h Monte Carlo simulation](https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/demo_synthetic_simulation.png)
+
+### 4. Real data — weekly backtest on DEIS
+
+The DEIS demo runs the same forecasting layer against real ED attendances from
+Hospital de Puerto Montt. The harness holds out the last 12 complete weeks
+(partial edge weeks are trimmed, since near-real-time DEIS data can under-count
+the latest week) and trains on the prior history; `AutoARIMA` is selected by P80
+pinball loss and tracks the autumn rise — evidence the pipeline isn't overfit to
+the synthetic regime.
+
+![12-week holdout, Puerto Montt](https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/deis_holdout_hospital_base_puerto_montt.png)
+
+### 5. Operational 6-month forecast
+
+With the model validated, the demo retrains on all history and emits a 26-week
+weekly forecast — the useful horizon for staffing, budget, and supplies. The
+P80–P95 band widens with the horizon, as expected.
+
+![6-month forecast, Puerto Montt](https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/deis_forecast_hospital_base_puerto_montt.png)
+
+### 6. Reference dashboard
+
+The FastAPI server (`urgencias-server`) exposes four routes with the same charts
+as the pipeline, live in the browser. It is deliberately minimal — Jinja2 plus
+base64-embedded matplotlib, no JavaScript — a starting point for a hospital to
+clone and adapt.
+
+<p align="center">
+<img src="https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/dashboard_index.png" width="48%" alt="Dashboard - index"/>
+<img src="https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/dashboard_baseline.png" width="48%" alt="Dashboard - descriptive analytics"/>
+</p>
+<p align="center">
+<img src="https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/dashboard_forecast.png" width="48%" alt="Dashboard - forecast"/>
+<img src="https://raw.githubusercontent.com/nicoveraz/urg-forecast-core/main/docs/img/dashboard_simulation.png" width="48%" alt="Dashboard - simulation"/>
+</p>
+
+## What's inside
+
+| Module | Purpose |
+|---|---|
+| `urgencias_core.data.loader` | Read a visit-level parquet and validate the schema. |
+| `urgencias_core.data.timeseries` | Convert visits to an hourly series (arrivals, discharges, occupancy by acuity, mean LOS) via the event-cumsum trick. |
+| `urgencias_core.data.deis` | DEIS MINSAL client (fetch + cache + filter to demo hospitals) with an offline snapshot fallback. |
+| `urgencias_core.features.calendar` | `holidays.CL` holidays, bridge days, school calendar, configurable regional events. |
+| `urgencias_core.features.weather` | Open-Meteo client with on-disk cache, Puerto Montt by default. |
+| `urgencias_core.models.protocol` | `Forecaster` protocol and `HorizonSpec` (grain-agnostic: hourly, daily, weekly, monthly). |
+| `urgencias_core.models.lgb_quantile` | LightGBM quantile regression, one model per quantile, calendar features. |
+| `urgencias_core.eval.baselines` | `SeasonalNaiveBaseline` + `statsforecast` wrappers (AutoARIMA, AutoETS, AutoTheta, MSTL). |
+| `urgencias_core.eval.harness` | Side-by-side evaluation with a ≥5% warning rule (a model that doesn't beat the baselines shouldn't ship). |
+| `urgencias_core.simulation.los_empirical` | Empirical LOS sampler conditional on (acuity, arrival hour). |
+| `urgencias_core.simulation.engine` | Monte Carlo census simulation 24 hours forward. |
+| `urgencias_core.server` | Minimal FastAPI server (4 routes, Jinja2, base64 matplotlib, no JS). |
+
+Docstrings are in English; user-facing strings in the server, demo outputs, and
+fixtures are in Spanish.
+
+## Public data: DEIS MINSAL
+
+The DEIS demo uses open data from Chile's Ministry of Health Department of Health
+Statistics and Information ([deis.minsal.cl](https://deis.minsal.cl/#datosabiertos))
+for two hospitals in the Servicio de Salud Reloncaví network: **Hospital de
+Puerto Montt** (code 24-105, high-complexity, locally "Hospital Base de Puerto
+Montt") and **Hospital de Frutillar** (code 24-115, low-complexity).
+
+DEIS publishes this series from 2008 to the present. The current-year file is
+updated weekly during the winter respiratory campaign (March–September) and
+roughly monthly otherwise. The demo uses the latest data available at run time
+and produces a six-month-ahead forecast.
+
+**Why these two hospitals.** Regional reference centers in Los Lagos with
+publicly available data. Chosen on geographic and pragmatic grounds, not on any
+evaluation of quality.
+
+**Methodological framing only.** This demo uses public DEIS MINSAL data to show
+how the forecasting tools behave on real Chilean hospital data. It is NOT an
+operational, clinical, or quality evaluation of either hospital.
+
+**Attribution and license.** Data published by DEIS MINSAL under Chile's open
+data framework. If you use this code or derivatives for research, maintain DEIS
+attribution and, where relevant, cite `urg-forecast-core`. The offline snapshot
+bundled with the package is a filtered extract of the public dataset for
+reproducibility; it does not replace fetching directly from the source for
+operational use.
+
+## Project status
+
+`urg-forecast-core` is an open foundation, developed and maintained by Nicolás
+Vera Z. as the base of **Eunosia**, a clinical AI platform for emergency
+medicine. It is published on PyPI and maintained on a best-effort basis: while
+on 0.x the API may change between minor versions, and issues and pull requests
+are welcome but not guaranteed a fast response. If you need commercial support
+or bespoke work on top of this foundation, contact the author.
+
+See [`CONTRIBUTING.md`](https://github.com/nicoveraz/urg-forecast-core/blob/main/CONTRIBUTING.md)
+for development and release procedures, [`docs/decisions.md`](https://github.com/nicoveraz/urg-forecast-core/blob/main/docs/decisions.md)
+for architecture decisions, and [`docs/roadmap.md`](https://github.com/nicoveraz/urg-forecast-core/blob/main/docs/roadmap.md)
+for deferred items (xlsx/mdb support for DEIS 2017–2019, neuralforecast, EMR
+integration to separate workup from boarding time).
+
+## Citing
+
+If you use `urg-forecast-core` in research, please cite it. Machine-readable
+metadata lives in [`CITATION.cff`](https://github.com/nicoveraz/urg-forecast-core/blob/main/CITATION.cff)
+(GitHub renders a "Cite this repository" button from it). Each tagged release is
+archived on [Zenodo](https://zenodo.org/) with a DOI.
+
+<!-- After the first Zenodo release, add the concept-DOI badge and BibTeX here:
+[![DOI](https://zenodo.org/badge/DOI/10.5281/zenodo.XXXXXXX.svg)](https://doi.org/10.5281/zenodo.XXXXXXX) -->
+
+A software paper (JOSS format) and a fuller methods preprint are drafted under
+[`paper/`](https://github.com/nicoveraz/urg-forecast-core/tree/main/paper).
+Please also mention Eunosia and link back to the repository.
+
+## License
+
+MIT. See [LICENSE](https://github.com/nicoveraz/urg-forecast-core/blob/main/LICENSE).