test_eval_chain.py — langchain Source File

Source Code

"""Test LLM Bash functionality."""

import os
import sys
from unittest.mock import patch

import pytest

from langchain_classic.chains.llm import LLMChain
from langchain_classic.evaluation.loading import load_evaluator
from langchain_classic.evaluation.qa.eval_chain import (
    ContextQAEvalChain,
    CotQAEvalChain,
    QAEvalChain,
    _parse_string_eval_output,
)
from langchain_classic.evaluation.schema import StringEvaluator
from tests.unit_tests.llms.fake_llm import FakeLLM


@pytest.mark.skipif(
    sys.platform.startswith("win"),
    reason="Test not supported on Windows",
)
def test_eval_chain() -> None:
    """Test a simple eval chain."""
    example = {"query": "What's my name", "answer": "John Doe"}
    prediction = {"result": "John Doe"}
    fake_qa_eval_chain = QAEvalChain.from_llm(FakeLLM())

    outputs = fake_qa_eval_chain.evaluate([example, example], [prediction, prediction])
    assert outputs[0] == outputs[1]
    assert fake_qa_eval_chain.output_key in outputs[0]
    assert outputs[0][fake_qa_eval_chain.output_key] == "foo"


@pytest.mark.skipif(
    sys.platform.startswith("win"),
    reason="Test not supported on Windows",
)
@pytest.mark.parametrize("chain_cls", [ContextQAEvalChain, CotQAEvalChain])
def test_context_eval_chain(chain_cls: type[ContextQAEvalChain]) -> None:
    """Test a simple eval chain."""
    example = {
        "query": "What's my name",
        "context": "The name of this person is John Doe",
    }
    prediction = {"result": "John Doe"}
    fake_qa_eval_chain = chain_cls.from_llm(FakeLLM())

    outputs = fake_qa_eval_chain.evaluate([example, example], [prediction, prediction])
    assert outputs[0] == outputs[1]
    assert "text" in outputs[0]
    assert outputs[0]["text"] == "foo"


def test_load_criteria_evaluator() -> None:
    """Test loading a criteria evaluator."""
    try:
        from langchain_openai import ChatOpenAI  # noqa: F401
// ... (93 more lines)