chore(llmobs): implement answer relevancy ragas metric

[lievan]

Implements answer relevancy metric for ragas integration.

About Answer Relevancy

Answer relevancy metric focuses on assessing how pertinent the generated answer is to the given prompt. A lower score is assigned to answers that are incomplete or contain redundant information and higher scores indicate better relevancy. This metric is computed using the question, the retrieved contexts and the answer.

The Answer Relevancy is defined as the mean cosine similarity of the original question to a number of artificial questions, which where generated (reverse engineered) based on the response.

Example trace

Checklist

• PR author has checked that all the criteria below are met
• The PR description includes an overview of the change
• The PR description articulates the motivation for the change
• The change includes tests OR the PR description describes a testing strategy
• The PR description notes risks associated with the change, if any
• Newly-added code is easy to change
• The change follows the library release note guidelines
• The change includes or references documentation updates if necessary
• Backport labels are set (if applicable)

Reviewer Checklist

• Reviewer has checked that all the criteria below are met
• Title is accurate
• All changes are related to the pull request's stated goal
• Avoids breaking API changes
• Testing strategy adequately addresses listed risks
• Newly-added code is easy to change
• Release note makes sense to a user of the library
• If necessary, author has acknowledged and discussed the performance implications of this PR as reported in the benchmarks PR comment
• Backport labels are set in a manner that is consistent with the release branch maintenance policy

[github-actions]

CODEOWNERS have been resolved as:

ddtrace/llmobs/_evaluators/ragas/answer_relevancy.py                    @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/tests.llmobs.test_llmobs_ragas_answer_relevancy_evaluator.emits_traces_and_evaluations_on_exit.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/tests.llmobs.test_llmobs_ragas_evaluators.test_ragas_answer_relevancy_emits_traces.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/tests.llmobs.test_llmobs_ragas_evaluators.test_ragas_answer_relevancy_submits_evaluation.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/tests.llmobs.test_llmobs_ragas_evaluators.test_ragas_answer_relevancy_submits_evaluation_on_span_with_custom_keys.yaml  @DataDog/ml-observability
tests/llmobs/llmobs_cassettes/tests.llmobs.test_llmobs_ragas_evaluators.test_ragas_answer_relevancy_submits_evaluation_on_span_with_question_in_messages.yaml  @DataDog/ml-observability
ddtrace/llmobs/_evaluators/ragas/base.py                                @DataDog/ml-observability
ddtrace/llmobs/_evaluators/ragas/models.py                              @DataDog/ml-observability
ddtrace/llmobs/_evaluators/runner.py                                    @DataDog/ml-observability
tests/llmobs/_utils.py                                                  @DataDog/ml-observability
tests/llmobs/conftest.py                                                @DataDog/ml-observability
tests/llmobs/test_llmobs_ragas_evaluators.py                            @DataDog/ml-observability

[datadog-dd-trace-py-rkomorn]

Datadog Report

Branch report: evan.li/answer-relevancy
Commit report: 812a162
Test service: dd-trace-py

✅ 0 Failed, 558 Passed, 910 Skipped, 7m 20.85s Total duration (28m 50.51s time saved)

[pr-commenter]

Benchmarks

Benchmark execution time: 2024-12-16 16:02:25

Comparing candidate commit 812a162 in PR branch evan.li/answer-relevancy with baseline commit 956e32d in branch evan.li/remaining-ragas-metrics.

Found 0 performance improvements and 0 performance regressions! Performance is the same for 394 metrics, 2 unstable metrics.

[Yun-Kim]

dumb question - what is the purpose of having a different LLM instance per eval metric runner? Are they not all references to the same base OpenAI() LLM?

[lievan]

Each eval metric runner has access to one instance of a ragas metric (answer_relevancy, context_precision, faithfulness) each of these ragas metrics have a seperate llm attribute. We maintain a reference to the ragas metric, not the llm attribute

but yea, the default is openai for all of them

[Yun-Kim]

Same comment as before, doesn't seem necessary to explicitly separate into new variables since they're only called once immediately at most

[Yun-Kim]

Why is this fmt comment required?

[lievan]

there was a error being raised by mypy due to the use of ABC. I think it was a bug since it only appeared when we have 3 or more evaluators (see: python/mypy#13044). But we are not using an abstract class for BaseRagasEvaluator anyway, so no need for this anymore

[lievan]

Going to re-open another PR (#11915) with latest changes from @main