Skip to content

feat: the Settings page says whether searches use the search questions - #97

Merged
alkhatibdev merged 1 commit into
mainfrom
search-questions-settings
Nov 10, 2026
Merged

alkhatibdev merged 1 commit into
mainfrom
search-questions-settings

Conversation

@alkhatibdev

Copy link
Copy Markdown
Owner

The search questions on the Settings page, and in the governance log

The Settings page shows how an instance is set up and changes nothing, and the search questions
are now part of it: whether searches use them, and for each law how many it has, how many are
current and when they were written, with the models that wrote and checked them. They are
turned off with dpolens search-questions off, which the governance log records, and the log
now reads its two new entries as sentences rather than as raw action names.

What changes

  • GET /v1/settings carries search_questions: on, and each law's slug, short name, how many
    questions it holds, how many are current, who wrote and checked them and when. It comes from
    the engine functions dpolens search-questions status uses. There is no write route: the
    page reads, and whoever runs the instance changes it, as with every other setting.
  • The Settings page has a Search questions group after AI provider. A law whose words changed
    under some of its questions says how many are current. The page's subtitle now says most
    settings are fixed when the instance starts, since this one is not.
  • The governance log reads search_questions.loaded as "Loaded 1,081 search questions for UAE
    PDPL, written 7 November 2026" and search_questions.switched as "Turned the search
    questions off" or "on". The loading entry now records its law in its details, as a pack's
    loading does.
  • docs/openapi.json, the API changelog's 0.3.0 list and the dashboard's generated types.

Checked

On 23347fa with this change, with GITHUB_ACTIONS=true, CI=true and every cache off:

  • ruff check, ruff format --check, mypy, lint-imports, scripts/api_contract.py,
    scripts/client_text.py, docker compose config
  • pytest: 1435 passed. At the database, GET /v1/settings with no questions loaded, and
    with a law's questions loaded, one written for words its clause no longer holds, and the
    switch off
  • The retrieval evals job: every published run at its published figure
  • The dashboard: prettier --check, eslint, tsc, 130 tests, and the production build
  • In Chromium and WebKit, at 1440 and 390 pixels wide, with the page's data served as in its
    tests: no horizontal overflow, no console errors, each model's name on one line, and the two
    laws' rows matching

@alkhatibdev
alkhatibdev merged commit 9fe00ea into main Nov 10, 2026
4 checks passed
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant