Implementation:Explodinggradients Ragas MetricResult Class

Field	Value
source	Explodinggradients_Ragas (GitHub)
domains	Metrics, Data_Model
last_updated	2026-02-10 00:00 GMT

Overview

The MetricResult class is a transparent wrapper around metric evaluation results that supports arithmetic operations, comparisons, iteration, and serialization while preserving access to metadata such as reasoning and traces.

Description

MetricResult stores an underlying value of any type along with optional reason (string) and traces (dictionary with "input"/"output" keys) metadata. The class implements Python's numeric dunder methods (__add__, __sub__, __mul__, __truediv__ and their reverse variants) for numeric results, container protocols (__getitem__, __iter__, __len__) for list results from ranking metrics, and type conversion methods (__float__, __int__) for numeric results. Comparison operators work across all value types. The __getattr__ fallback forwards attribute access to the underlying value, wrapping returned values of the same type back in MetricResult to preserve metadata. Pydantic integration is provided through __get_pydantic_core_schema__ for validation and serialization.

Usage

MetricResult instances are typically created by metric scoring methods. Use the .value property to access the raw result, .reason for the LLM reasoning, and .to_dict() or .__json__() for serialization.

Code Reference

Item	Detail
Source Location	`src/ragas/metrics/result.py` L11-241
Class Signature	`class MetricResult`
Constructor	`def __init__(self, value: Any, reason: Optional[str] = None, traces: Optional[Dict[str, Any]] = None)`
Import	`from ragas.metrics.result import MetricResult`

I/O Contract

Inputs

Parameter	Type	Description
`value`	`Any`	The raw metric result (float, str, list, etc.)
`reason`	`Optional[str]`	Optional reasoning text from the LLM
`traces`	`Optional[Dict[str, Any]]`	Optional traces dict; only "input" and "output" keys are allowed

Outputs

Property/Method	Type	Description
`.value`	`Any`	The raw underlying result value
`.reason`	`Optional[str]`	The reasoning metadata
`.to_dict()`	`Dict`	Dictionary with "result" and "reason" keys
`.__json__()`	`Dict`	Dictionary with "value" and "reason" keys for JSON serialization

Usage Examples

from ragas.metrics.result import MetricResult

# Numeric result with arithmetic
result = MetricResult(value=0.85, reason="High faithfulness")
print(float(result))       # 0.85
print(result + 0.1)        # 0.95
print(result > 0.5)        # True

# List result for ranking metrics
ranking = MetricResult(value=["doc_a", "doc_b", "doc_c"])
print(len(ranking))        # 3
print(ranking[0])          # 'doc_a'
for item in ranking:
    print(item)

# Serialization
print(result.to_dict())    # {'result': 0.85, 'reason': 'High faithfulness'}

Related Pages

RankingMetric_Class - Produces list-valued MetricResult instances
MetricValidators_Module - Validates MetricResult values against constraints
Utils_Module - safe_nanmean works with MetricResult float values

Page Connections

Double-click a node to navigate. Hold to expand connections.

Principle

Implementation

Heuristic

Environment