Evaluator

The Evaluator node uses AI to score content on a scale of 0-100 based on custom criteria you define. It's ideal for quality assessment, relevance scoring, and content evaluation.

Overview

Use the Evaluator node when you need to:

Score content quality or relevance
Assess how well content meets specific criteria
Rank or prioritize items based on scores
Filter content based on quality thresholds

Configuration

Field	Type	Required	Description
`model`	`string`	Yes	The AI model to use for evaluation.
`criteria`	`string`	Yes	The scoring criteria. Be specific about what constitutes high and low scores.
`includeJustification`	`boolean`	No	When enabled, returns an explanation for the score. Default: `false`.

Loop Mode

The Evaluator node supports Loop Mode for batch evaluation:

Field	Type	Default	Description
`loopMode`	`boolean`	`false`	Enable to evaluate each item in an array input separately.
`maxIterations`	`number`	`100`	Maximum number of iterations when loop mode is enabled.
`concurrency`	`number`	`1`	Number of parallel evaluations.
`onError`	`string`	`"stop"`	Error handling: `"stop"` or `"continue"`.

Inputs

The Evaluator node accepts inputs from connected upstream nodes. All connected inputs are automatically concatenated and evaluated together.

Output

Variable	Type	Description
`score`	`number`	A score from 0-100 based on the criteria.
`justification`	`string`	(Optional) Explanation for the score. Only present when `includeJustification` is enabled.

Example Use Cases

Content Quality Assessment

Criteria:

Score based on:
- Clarity (0-30): Is the content easy to understand?
- Relevance (0-40): Does it address the topic appropriately?
- Accuracy (0-30): Is the information correct and well-supported?

Input:

Machine learning is a subset of artificial intelligence that enables systems to learn from data...

Output:

{
  "score": 85,
  "justification": "The content is clear and well-written (28/30), highly relevant to the topic (38/40), and accurate with good explanations (19/30)."
}

Email Response Quality

Criteria:

Evaluate the email response on:
- Professionalism (0-25): Appropriate tone and language
- Completeness (0-35): Addresses all questions asked
- Helpfulness (0-25): Provides actionable information
- Clarity (0-15): Easy to understand and follow

Lead Scoring

Criteria:

Score the lead based on:
- Intent signals (0-40): Keywords indicating purchase intent
- Company fit (0-30): Company size, industry, and relevance
- Engagement level (0-30): Interaction history and responsiveness

Resume Screening

Criteria:

Evaluate the resume for a Software Engineer position:
- Technical skills match (0-35): Relevant programming languages and technologies
- Experience level (0-30): Years and quality of relevant experience
- Education (0-15): Relevant degrees and certifications
- Communication (0-20): Clarity and presentation of information

Best Practices

Define clear point allocations: Break down your criteria into components with specific point values that add up to 100.
Be specific about thresholds: Describe what constitutes a high score (80+), medium score (50-79), and low score (0-49) for your use case.
Include examples in criteria: When possible, provide examples of what would score high or low.
Enable justification for transparency: The justification helps you understand and validate the scoring logic.
Use consistent criteria: For comparable scores across different inputs, keep the criteria consistent.
Calibrate with test data: Run several examples through the evaluator to ensure scores align with your expectations.

Combining with Other Nodes

The Evaluator node works well with:

Router: Use the score to route content (e.g., scores > 80 go to "approved", < 50 go to "review")
Classifier: Combine evaluation with classification for multi-dimensional analysis
Extract Data: Evaluate extracted information quality

Evaluator

On this page