Implement SRP Complexity Classifier with Multi-language Support

[JeremyDev87]

Implement SRP Complexity Classifier with Multi-language Support

Summary

This PR introduces an intelligent complexity classification system for the PLAN mode workflow. The classifier analyzes user prompts to determine whether they require structured reasoning (SRP) or can be handled directly, supporting 5 languages with negation pattern detection.

Implementation Details

1. Complexity Classification Algorithm

Score = Σ(matched_indicator.weight) / Σ(all_indicator.weight)

If COMPLEX_score > 0.08 AND COMPLEX_score > SIMPLE_score → COMPLEX
If SIMPLE_score > 0.10 AND SIMPLE_score > COMPLEX_score → SIMPLE
Otherwise → Borderline handling with safety margin

2. Negation Pattern Detection

Implemented language-aware negation detection:

• Prefix negations (EN, ZH, ES): "don't", "不要", "no"
• Suffix negations (KO, JA): "하지 말고", "しないで"

The isNegatedMatch() function checks 30 characters before and 15 characters after each match to detect negation context.

3. Function Decomposition

Refactored classifyComplexity() into focused helpers:

classifyComplexity(prompt, options)
  ├── extractSrpOverride(prompt)      // Handle --srp/--no-srp flags
  ├── calculateScores(prompt)          // Score both indicator types
  │     ├── calculateScore(COMPLEX_INDICATORS, checkNegation=true)
  │     └── calculateScore(SIMPLE_INDICATORS, checkNegation=false)
  ├── determineComplexity(scores)      // Apply thresholds
  └── applyOverride(result, override)  // Handle user overrides

4. Multi-language Support

Language	COMPLEX Keywords	SIMPLE Keywords	Negation Patterns
English	design, architect, refactor, optimize	fix, format, typo, comment	don't, do not, skip, avoid
Korean	설계, 아키텍처, 리팩토링, 최적화	수정, 포맷, 오타, 주석	하지 말, 안 해도, 필요 없
Japanese	設計, アーキテクチャ, リファクタリング	修正, フォーマット, タイポ	しないで, 必要ない
Chinese	设计, 架构, 重构, 优化	修复, 格式化, 错字	不要, 不需要, 别
Spanish	diseño, arquitectura, refactorizar	corregir, formatear, error	no, sin, evitar

Test Coverage

• 94 tests covering all scenarios
• 13ms total test execution time
• Test categories:
  • Basic classification (COMPLEX/SIMPLE)
  • Override handling (--srp, --no-srp)
  • i18n patterns (EN, KO, JA, ZH, ES)
  • Negation detection (21 tests)
  • Mixed-language prompts (6 tests)
  • Telemetry callback
  • Configurable thresholds
  • Edge cases and borderline scenarios

Quality Assurance

Code Quality Review: LOW RISK

• Excellent SOLID compliance
• Clean function decomposition
• Strong type safety (no any)
• Backwards compatibility maintained

Performance Review: LOW RISK

• Bundle size: ~27KB (acceptable for MCP server)
• Test execution: 13ms
• No critical performance issues

Breaking Changes

None. The API is additive and backwards compatible.

How to Test

# Run complexity classifier tests
yarn workspace codingbuddy test src/keyword/complexity-classifier.spec.ts

# Run all keyword module tests
yarn workspace codingbuddy test src/keyword/

Related

close #274