§ Evaluation · Custom Metrics Configuration
Custom Evaluation Metrics
Define custom validation rules using regex patterns, text matching, or model-based validators to be applied during golden dataset testing.
Total Metrics
0
Active
0
Metric Types
0
Configured Metrics
No custom metrics yet
Create your first metric to enhance evaluation coverage
📋 Metric Types:
Regex: Match patterns in output (PII detection, format validation)
Contains: Check if output includes specific text
Length Check: Validate output length constraints
Model Validator: Use LLM to validate semantic properties (sentiment, tone, safety)
PII Detection
Type: regex
JSON Format
Type: regex
Sentiment Check
Type: model_validator
Output Length
Type: length_check