feat: make general purpose metrics more general

[jjmachan]

Metrics Converted

• [x] Aspect Critic
• [x] Simple Criteria
• [x] Rubric Based - both Instance and Domain specific

a few different examples

Aspect Critic

from ragas.metrics import AspectCritic
from ragas.dataset_schema import SingleTurnSample

only_response = SingleTurnSample(
    response="The Eiffel Tower is located in Paris."
)

grammar_critic = AspectCritic(
    name="grammar",
    definition="Is the response grammatically correct?",
    llm=evaluator_llm
)

await grammar_critic.single_turn_ascore(only_response)

with reference

answer_correctness_critic = AspectCritic(
    name="answer_correctness",
    definition="Is the response and reference answer are the same?",
    llm=evaluator_llm
)

# data row
sample = SingleTurnSample(
    user_input="Where is the Eiffel Tower located?",
    response="The Eiffel Tower is located in Paris.",
    reference="London"
)
await answer_correctness_critic.single_turn_ascore(sample)

Note: this only works for multi-turn metrics for now

[shahules786]

• Let's say I am evaluating using an evaluation dataset that contains user_input, response, reference, and retrieved_context. Consider scenarios of the user using two metrics with this dataset, ie context_recall and aspect critic (as harmfulness or something) at the same time. Using this interface, even if my aspect critic metric does not need retrieved context it will use it. There is no way to opt-out. I think this can occur when using the metrics using the evaluate interface.

• When using a single metric this seems to be fine. @jjmachan