EvalResult

class EvalResult:
    input: Any              # Required: test input
    output: Any             # Required: system output
    reference: Any = None   # Optional: expected output
    scores: list[Score] = []  # Optional: list of scores
    error: str = None       # Optional: error message
    latency: float = None   # Optional: execution time in seconds
    metadata: dict = {}     # Optional: custom metadata
    trace_data: dict = {}     # Optional: debug/trace data

EvalResult(
    input="What is the capital of France?",
    output="Paris"
)

EvalResult(
    input="2 + 2",
    output={"answer": 4, "confidence": 0.99}
)

EvalResult(
    input="What is 2 + 2?",
    output="4",
    reference="4"
)

EvalResult(
    input="test",
    output="result",
    scores=[
        {"key": "accuracy", "passed": True},
        {"key": "latency", "value": 0.95}
    ]
)

EvalResult(
    input="test",
    output=None,
    error="TimeoutError: Evaluation exceeded 5.0 seconds"
)

EvalResult(
    input="test",
    output="result",
    latency=0.234
)

EvalResult(
    input="test",
    output="result",
    metadata={
        "model": "gpt-4",
        "temperature": 0.7,
        "category": "factual"
    }
)

EvalResult(
    input="test",
    output="result",
    trace_data={
        "tokens_used": 150,
        "api_calls": 2,
        "trace_id": "abc123"
    }
)

@eval(dataset="demo")
async def my_eval(ctx: EvalContext):
    ctx.input = "test"
    ctx.output = await my_agent(ctx.input)

    assert ctx.output, "Got empty response"
    # Returns EvalResult automatically

from ezvals import EvalResult

result = EvalResult(
    input="What is 2 + 2?",
    output="4",
    reference="4",
    scores=[{"key": "accuracy", "passed": True}],
    metadata={"difficulty": "easy"}
)

@eval(dataset="batch")
def test_batch():
    return [
        EvalResult(input="a", output="1"),
        EvalResult(input="b", output="2"),
        EvalResult(input="c", output="3"),
    ]

result = EvalResult(input="test", output="result")

print(result.input)    # "test"
print(result.output)   # "result"
print(result.latency)  # None (not set)

result = EvalResult(
    input="test",
    output="result",
    scores=[{"key": "accuracy", "passed": True}]
)

# To dict
data = result.model_dump()

# To JSON string
json_str = result.model_dump_json()

# Passed - has a passing score
EvalResult(input="x", output="y", scores=[{"key": "a", "passed": True}])

# Passed - mixed scores, but at least one passes
EvalResult(input="x", output="y", scores=[
    {"key": "a", "passed": True},
    {"key": "b", "passed": False}
])

# Failed - all scores fail
EvalResult(input="x", output="y", scores=[{"key": "a", "passed": False}])

# Failed - has error
EvalResult(input="x", output="y", error="Something went wrong")

from ezvals import EvalResult

result = EvalResult(
    input={
        "question": "What is the capital of France?",
        "context": "Geography quiz"
    },
    output="The capital of France is Paris.",
    reference="Paris",
    scores=[
        {
            "key": "accuracy",
            "passed": True,
            "notes": "Correct answer"
        },
        {
            "key": "brevity",
            "value": 0.7,
            "notes": "Could be more concise"
        }
    ],
    latency=0.234,
    metadata={
        "model": "gpt-4",
        "temperature": 0.0,
        "category": "geography"
    },
    trace_data={
        "tokens": 45,
        "api_latency": 0.189
    }
)