WIP LLM ranking/scoring backend

[juhoinkinen]

Closes #856.

[codecov]

Codecov Report

❌ Patch coverage is 1.87500% with 157 lines in your changes missing coverage. Please review.
✅ Project coverage is 97.56%. Comparing base (6bae2e5) to head (2013e9c).
⚠️ Report is 196 commits behind head on main.

Files with missing lines	Patch %	Lines
annif/backend/llm_ensemble.py	0.00%	155 Missing ⚠️
annif/backend/__init__.py	33.33%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main     #859      +/-   ##
==========================================
- Coverage   99.64%   97.56%   -2.09%     
==========================================
  Files          99      100       +1     
  Lines        7349     7509     +160     
==========================================
+ Hits         7323     7326       +3     
- Misses         26      183     +157     

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[Copilot]

Pull Request Overview

This PR adds exponentiated weighted averaging to suggestions and implements an LLM-based ensemble backend for ranking and scoring.

• Extend SuggestionBatch.from_averaged to accept an optional exponents parameter for score exponentiation.
• Introduce BaseLLMBackend and LLMEnsembleBackend with OpenAI/AzureOpenAI integration and parallel prompt processing.
• Register the new llm_ensemble backend in the backend factory.

Reviewed Changes

Copilot reviewed 3 out of 3 changed files in this pull request and generated 3 comments.

File	Description
annif/suggestion.py	Added exponents parameter and updated averaging logic/docstring.
annif/backend/llm_ensemble.py	New LLM ensemble backend: API calls, prompt handling, ensemble logic.
annif/backend/init.py	Registered llm_ensemble backend.

Comments suppressed due to low confidence (2)

annif/suggestion.py:125

• [nitpick] Update the docstring for from_averaged to include a description of the new exponents parameter and its default behavior.

        """Create a new SuggestionBatch where the subject scores are the

annif/backend/llm_ensemble.py:263

• [nitpick] Add a brief docstring to _get_labels_batch to clarify its behavior and inputs, improving code readability.

    def _get_labels_batch(self, suggestion_batch: SuggestionBatch) -> list[list[str]]:

[sonarqubecloud]

Quality Gate passed

Issues
0 New issues
0 Accepted issues

Measures
0 Security Hotspots
0.0% Coverage on New Code
0.0% Duplication on New Code

See analysis details on SonarQube Cloud

[juhoinkinen]

Remainder for implementation: We could check if querying the LLM multiple times and averaging the results gives an improvement (this could work when using a non-zero temperature).

This approach is used in https://kata.rara.ee/ service.