Update version.json

…image automatically

…tch)

Small packages upgrade

Paginate the competition list endpoint (LargePagination), and update
the Organizing/Participating tabs in competition_list.tag to
consume the new {count, next, previous, results} shape with
next/previous buttons (hidden on a single page). Fix the navbar
competition search to read response.results instead of assuming a
bare array.

Fix participating_in filter incorrectly including competitions the
user organizes: creators/collaborators are auto-added as approved
participants (Competition.save()), so they leaked into the
Participating tab. Applied the same fix to the public competitions
endpoint's participating_in filter and updated its tests. Added
coverage for the list endpoint's pagination and participating_in
behavior. Minor renames for clarity (change_page, tab labels).

Add pagination to GET /api/competitions and fix participating_in leak

…limit

Raise RAM limit for Site Worker in docker-compose.yml

…t is reached. Fixed user lookup to not show deleted users in the collaborators dropdown

Queue creation error improved

Remove the is_bot user flag and allow_robot_submissions competition setting, along with the permission bypasses they granted (unlimited submissions, auto-approval as participant, and unrestricted re-run/secret-check bypass on any submission). Also removes the related UI, docs, and tests.

Remove robot submissions and bot user feature

Fix Django migrations

Privacy link updated

…titions

Propagate organization when creating chiled submissions during multi-task split, phase migration, and re-run so that leaderboard entries for organization submissions on multi-task competitions show the organization name instead of falling back to the submitter's username.

Fix organization name not showing on leaderboard for multi-task competitions

API Throttling

Anonymous and non-owner requests could previously view submission
details when hide_output was off. Now only the owner or an admin
can access them. Adds tests covering the fix.

Submission `get_detail` API: Restrict to submission owner or admin

* Deny anonymous access to /api/submissions/ list and retrieve

Anonymous GETs returned full submission records (filenames, status
details, fact sheet answers) for any finished leaderboard submission,
ignoring is_public. Now returns an empty queryset instead; public
leaderboard data is already served by get_leaderboard in PhaseViewSet

* Add missing raise

* Fix test and remove debug prints

---------

Co-authored-by: didayolo <adrien.pavao@gmail.com>

…#2496)

* Enforce phase start/end window on submission creation

Phase.is_active returned True unconditionally when end was unset,
never checking start — a not-yet-started phase with no end date was
treated as active. SubmissionCreationSerializer.validate() also never
checked is_active or can_user_make_submissions(), so submissions could
be created via the API before a phase started or after it ended.

- Fix Phase.is_active to check start regardless of whether end is set
- Reject submission creation when the target phase is not active or
  the user has hit their submission limit, returning a 400 error
- Add tests for both new fixes

* Fix flake8

---------

Co-authored-by: didayolo <adrien.pavao@gmail.com>

* Add External Competitions feature: aggregate competitions from other Codabench/CodaLab instances

New app — external_competitions
- Models: ExternalPlatform (a registered remote instance), ExternalCompetition (a fetched competition), ExternalFetchLog (per-sync audit trail with SUCCESS/PARTIAL_SUCCESS/FAILURE status, counts, and error messages)
- admin.py registrations for all three models with list filters/search
- Initial migration

Fetchers
- fetchers/codabench_fetcher.py — paginated fetch from a Codabench instance's public competitions API, with a page-fetch throttle and a MAX_PAGES safety cap
- fetchers/codalab_fetcher.py — flat-list fetch from a CodaLab instance
- fetchers/exceptions.py — PartialFetchError, letting a fetcher hand back whatever it collected before a later page failed

Sync task
- fetch_sync.py — fetch_external_competitions Celery task (feature-flag gated, scheduled daily via CELERY_BEAT_SCHEDULE) and sync_platform, which diffs fetched data against the DB (create/update/delete), skipping the delete step on a partial fetch to avoid dropping valid competitions from unfetched pages

API
- api/serializers/external_competitions.py — ExternalCompetitionSerializer, ExternalPlatformFilterSerializer
- api/views/external_competitions.py — list view with search/platform filtering, and an unpaginated platforms list for the filter UI
- api/urls.py — both routes registered only when EXTERNAL_COMPETITIONS_ENABLED is on

Frontend
- external_competitions/urls.py + views.py + templates/external_competitions/public.html — the public page route
- static/riot/external_competitions/external_competition_list.tag — the competition list/filter/pagination UI
- client.js — API helpers for the two new endpoints
- public-list.tag — banner linking to the External Competitions page (shown only when the feature is enabled)
- base.html / context_processors.py — expose the feature flag and public URL to the frontend

Settings
- EXTERNAL_COMPETITIONS_ENABLED flag (settings/base.py, .env_sample), off by default
- external_competitions added to INSTALLED_APPS

Test data
- factories.py — ExternalPlatformFactory, ExternalCompetitionFactory

Tests
- external_competitions/tests/test_fetchers.py — unit tests for both fetchers (pagination, throttling, MAX_PAGES, error propagation, partial-fetch handling)
- external_competitions/tests/test_fetch_sync.py — unit tests for sync_platform/fetch_external_competitions (create/update/delete diffing, partial-success handling, per-platform failure isolation)
- api/tests/test_external_competitions.py — API tests for both endpoints, plus flag-gated URL registration tests

Documentation
- New External-Competitions.md page: enabling the flag, admin fields, sync mechanics, and registration/unregistration instructions for other platform maintainers
- zensical.toml nav entry
- Tip added to the deploy guide pointing self-hosters at registering their instance

* flake-8 fixes

* registration instructions added to external competitions page

* renamed fetch_sync.py to tasks.py so that celery can run the task periodically, updated codabench fetcher to not use the next url from the response but instead use a page counter for next page, related updates to docs and tests

* ui updates to differentiate external page from public page.

* Some small fixes

* Fix spinner

* Avoid duplicated CSS

---------

Co-authored-by: didayolo <adrien.pavao@gmail.com>

Compute worker workaround for API throttling

…es to Codabench (#2460)

* add option in compute worker env to not send logs to the instance, instead writing them in a local file

* rename the No Cleanup env variable, add documentation

* use real boolean values

* update logs_loguru to inclue new tasks variable names to color them

* change variable name to use boolean

* fix some syntax

* add option to forbid the compute worker from sending predictions to the codabench instance storage

* add better coloration for some logs in the compute worker

* update documentation

* add error when copying failes when in no prediction upload mode

---------

Co-authored-by: Obada Haddad <obada.haddad@lisn.fr>

Revert "Compute worker workaround for API throttling"

This reverts commit 17416df.

Revert "API Throttling"