feat: add multi-dimensional quality scorer for structured outputs

[lustsazeus-lab]

Summary

Implements a multi-dimensional quality scorer for structured outputs (JSON, markdown, code, text) as specified in issue #1.

Features

• 5-Dimension Scoring:

  • Completeness (30%): How complete the submission is
  • Format Compliance (20%): Follows expected format
  • Coverage (25%): Covers required aspects
  • Clarity (15%): Readable and clear content
  • Validity (10%): Valid and accurate content

• Auto-format Detection: Detects JSON, Markdown, Code, Text automatically

• Output Schema:

{
  "weighted_score": 0.85,
  "quality_rating": "excellent",
  "scores": {
    "completeness": 0.9,
    "format_compliance": 1.0,
    "coverage": 0.8,
    "clarity": 0.7,
    "validity": 0.9
  },
  "feedback": [...],
  "pass_threshold": true
}

• Performance: Batch scoring supports 100+ submissions efficiently

Tests

28 comprehensive tests covering:

• Format detection
• Individual dimension scoring
• Integration tests
• Edge cases

Usage

from scorer import QualityScorer

scorer = QualityScorer()
result = scorer.score(your_content)

print(f"Score: {result.weighted_score}")
print(f"Rating: {result.quality_rating.value}")

Bounty: (+ potential bonus for NLP feedback)