Upgrading from 1.x

Version 2.0.0 stopped scraping the ecolle website and started reading a CSV export instead. If you were running 1.x, here is what changes.

What changed

	1.x	2.x
Source of the colles	logged in to ecolle and scraped the agenda page	a CSV file you export yourself
BASE_URL, COLLES_USERNAME, COLLES_PASSWORD	required	only used by the scraper fallback, which is off by default
COLLES_CSV_PATH	did not exist	path of the export, default input/colles.csv
Container	had to be run on a machine that could reach ecolle	only needs the CSV file

The old variables are still read, so an existing .env does not break. They are simply no longer needed, and you should remove them so that a typo can no longer send the tool looking for a website:

-BASE_URL=https://ecolle.example.com
-COLLES_USERNAME=you
-COLLES_PASSWORD=secret
+COLLES_CSV_PATH=input/colles.csv
+DISABLE_ECOLLE_FETCH=true

Leaving them in place is harmless, but if the CSV file is ever missing the tool will try to scrape instead of telling you the file is missing.

Steps

1. Export your colles from ecolle as a CSV file, see Getting started.
2. Put it at ./input/colles.csv, next to your docker-compose.yml.
3. Add the two settings above to .env.
4. Make sure the input volume is mounted. The 1.x docker-compose.yml did not have it:

yaml volumes: - ./input:/app/input:ro - ./output:/app/output

1. Pull the new image and run it:

docker compose pull docker compose run --rm app

1. Check the first line of the output says [*] Reading CSV: input/colles.csv. If it says [!] No CSV found, see Troubleshooting.

About the notification wording

The default message changed in 2.3.0 to use French dates and times:

1.x:  Mathématiques 2026-10-06 16:00 L005 M Jouve
2.x:  mardi 6 octobre 16h00 L005 - Mathématiques (M Jouve)

If you had set NTFY_FORMAT yourself, your value is still used unchanged. If you were relying on the old default, either accept the new wording or pin it explicitly:

NTFY_FORMAT={matiere} {date} {heure} {salle} {colleur}

Versions in between

Version	Change
2.0.0	colles read from a CSV export, pandas dropped
2.1.0	DISABLE_ECOLLE_FETCH, clearer CSV errors
2.1.1	DISABLE_ECOLLE_FETCH spelling fixed (the misspelt name still works)
2.2.0	better default notification format
2.3.0	French dates and times
2.4.0	one shared date parser, cache no longer lost on a failed scrape, UTF-16 and encoding errors reported, container timezone fixed

The container image is tagged for each release (2.4.0, 2.4, 2), so pinning ghcr.io/florianzarbo/ecolle-scraper:2.4 avoids a surprise upgrade. latest follows the newest release.