The plugin follows modern SearXNG plugin architecture with SOLID principles, protocol-based design, and comprehensive unit testing.

I've only had a quick look at it so far, but I can already say that it's great work. Thank you very much 👍

Within the review of PR #3837, one important question has not yet been answered: we're not sure about the memory footprint of using pyspellchecker

• Add "Spelling / Did you mean?" feature #3837 (comment)

Consider adding .pre-commit-config.yaml to run these lints/CI cmds. I had no idea about them; my usual workflow is ruff and uv. Also I don't quite agree with a lot of these lint rules, e.g.

EDIT: I see the pr #3761 (comment) and the issue #3760 Hard disagree on the final decision. But okay.

Something like this...

fail_fast: true
default_install_hook_types: [pre-commit, pre-push]

repos:
  # General hooks
  - repo: https://github.com/pre-commit/pre-commit-hooks
    rev: v6.0.0
    hooks:
      - id: trailing-whitespace
      - id: end-of-file-fixer
      - id: check-yaml
      - id: check-json
      - id: check-merge-conflict
      - id: check-added-large-files
        args: ['--maxkb=5000']
      - id: check-case-conflict
      - id: check-merge-conflict
      - id: mixed-line-ending
        args: ['--fix=lf']
  - repo: local
    hooks:
      - id: format-python
        name: format Python code
        entry: ./manage format.python
        language: system
        types: [python]
        pass_filenames: false
        always_run: true
      - id: test-pylint
        name: run pylint
        entry: ./manage test.pylint
        language: system
        types: [python]
        pass_filenames: false
        always_run: true
  - repo: local
    hooks:
      - id: nose2
        name: run nose2 tests
        entry: uv run nose2 -F tests.unit
        # tests donn't work with mp plugin
        # entry: uv run --frozen nose2 -qq -F --plugin=nose2.plugins.mp --processes=$(nproc) tests.unit
        language: python
        types: [python]
        pass_filenames: false
        always_run: true
        stages: [pre-push]  # takes a long time, only run on pre-push

Amazing work! I've tested it with very good results.

we're not sure about the memory footprint of using pyspellchecker

It adds at most (based on my "trust me bro" tests) 20 MB additional memory usage with pyspellchecker enabled.

The issue here is that this library is CPU intensive, should be caped to 80 characters per search as it overloads the synchronous nature of WSGI, making the experience with concurrent requests very poor.

The issue here is that this library is CPU intensive,

Memory and/or CPU intensive tasks -- resource intensive tasks -- should not be performed by the SearXNG server. SearXNG is an aggregator and such tasks should be provided by other services to integrate them into SearXNG.

For pyspellchecker, it would be possible (not part of SearXNG) to implement a (local) service that is integrated into SearXNG via HTTP (like other engines, answerer or plugin).

For pyspellchecker, it would be possible (not part of SearXNG) to implement a (local) service that is integrated into SearXNG via HTTP (like other engines, answerer or plugin).

Working on a solution