breakpilot-compliance/ai-compliance-sdk/internal/usecase/compiler.go

package usecase

import (
	"fmt"
	"strings"

	"golang.org/x/text/cases"
	"golang.org/x/text/language"
)

// Compiler turns Master Controls into audit questionnaires.
type Compiler struct {
	store *Store
}

// NewCompiler creates a Compiler.
func NewCompiler(store *Store) *Compiler {
	return &Compiler{store: store}
}

// Compile generates questions for a template by combining pre-defined
// questions, existing doc_check_controls, and MC-derived questions.
func (c *Compiler) Compile(tmpl *Template) ([]Question, error) {
	// 1. Start with pre-defined template questions
	if len(tmpl.Questions) > 0 {
		return c.enrichWithMCIDs(tmpl)
	}

	// 2. Fetch MCs matching the template filters
	mcs, err := c.store.FetchMCsByFilters(tmpl.MCFilters)
	if err != nil {
		return nil, fmt.Errorf("fetch MCs: %w", err)
	}
	if len(mcs) == 0 {
		return nil, fmt.Errorf("no Master Controls found for filters %v", tmpl.MCFilters)
	}

	// 3. Check for existing doc_check_controls questions
	mcIDs := make([]string, len(mcs))
	for i, mc := range mcs {
		mcIDs[i] = mc.MasterControlID
	}

	checkQuestions, err := c.store.FetchCheckQuestions(mcIDs)
	if err != nil {
		return nil, fmt.Errorf("fetch check questions: %w", err)
	}

	// 4. Generate questions from MCs
	var questions []Question
	qNum := 1

	for _, mc := range mcs {
		// Mode A: Use existing doc_check questions
		if cqs, ok := checkQuestions[mc.MasterControlID]; ok {
			for _, cq := range cqs {
				q := Question{
					ID:           fmt.Sprintf("Q%d", qNum),
					MCID:         mc.MasterControlID,
					MCName:       mc.CanonicalName,
					Text:         cq.Question,
					QuestionType: "yes_no",
					Severity:     normalizeSeverity(cq.Severity),
					Regulation:   mc.RegSource,
					PassCriteria: splitCriteria(cq.PassCriteria),
					FailCriteria: splitCriteria(cq.FailCriteria),
				}
				questions = append(questions, q)
				qNum++
			}
			continue
		}

		// Mode A fallback: Derive question from MC name
		q := Question{
			ID:           fmt.Sprintf("Q%d", qNum),
			MCID:         mc.MasterControlID,
			MCName:       mc.CanonicalName,
			Text:         deriveQuestion(mc.CanonicalName),
			QuestionType: "yes_no",
			Severity:     inferMCSeverity(mc.CanonicalName),
			Regulation:   mc.RegSource,
			PassCriteria: []string{"Anforderung erfuellt und dokumentiert"},
			FailCriteria: []string{"Nicht implementiert oder nicht nachweisbar"},
		}
		questions = append(questions, q)
		qNum++

		// Cap at a reasonable number
		if qNum > 50 {
			break
		}
	}

	return questions, nil
}

// enrichWithMCIDs links pre-defined questions to MCs.
func (c *Compiler) enrichWithMCIDs(tmpl *Template) ([]Question, error) {
	mcs, err := c.store.FetchMCsByFilters(tmpl.MCFilters)
	if err != nil {
		return tmpl.Questions, nil // fallback to questions without MC linkage
	}

	mcByTopic := make(map[string]MCInfo)
	for _, mc := range mcs {
		mcByTopic[mc.CanonicalName] = mc
	}

	questions := make([]Question, len(tmpl.Questions))
	copy(questions, tmpl.Questions)

	// Try to link questions to MCs by keyword matching
	for i := range questions {
		if questions[i].MCID != "" {
			continue
		}
		qLower := strings.ToLower(questions[i].Text)
		for _, mc := range mcs {
			topic := strings.ReplaceAll(mc.CanonicalName, "_", " ")
			words := strings.Fields(topic)
			matched := 0
			for _, w := range words {
				if strings.Contains(qLower, w) {
					matched++
				}
			}
			if matched >= 2 {
				questions[i].MCID = mc.MasterControlID
				questions[i].MCName = mc.CanonicalName
				break
			}
		}
	}

	return questions, nil
}

// deriveQuestion generates a human-readable question from an MC name.
func deriveQuestion(canonicalName string) string {
	readable := strings.ReplaceAll(canonicalName, "_", " ")
	readable = cases.Title(language.German).String(readable)
	return fmt.Sprintf("Ist '%s' implementiert und dokumentiert?", readable)
}

// splitCriteria splits a pipe-separated criteria string.
func splitCriteria(s string) []string {
	if s == "" {
		return nil
	}
	parts := strings.Split(s, "|")
	result := make([]string, 0, len(parts))
	for _, p := range parts {
		p = strings.TrimSpace(p)
		if p != "" {
			result = append(result, p)
		}
	}
	if len(result) == 0 {
		return []string{s}
	}
	return result
}

// normalizeSeverity maps doc_check severity to our format.
func normalizeSeverity(s string) string {
	s = strings.ToUpper(strings.TrimSpace(s))
	switch s {
	case "HIGH", "CRITICAL":
		return "HIGH"
	case "MEDIUM":
		return "MEDIUM"
	case "LOW":
		return "LOW"
	default:
		return "MEDIUM"
	}
}

// inferMCSeverity guesses severity from the MC topic name.
func inferMCSeverity(name string) string {
	high := []string{"encryption", "access_control", "incident", "vulnerability",
		"authentication", "key_management", "data_breach", "personal_data",
		"consent", "data_transfer"}
	for _, h := range high {
		if strings.Contains(name, h) {
			return "HIGH"
		}
	}
	return "MEDIUM"
}