app/services/scoring.py

"""Scoring and decision logic with calibrated business rules.

Weights and decision thresholds default to the calibrated values below but can
be overridden at runtime by an administrator through the admin settings
("Configurer les paramètres du pipeline IA"). The current values are read from
``app.core.settings_store`` which falls back to these defaults when nothing has
been configured.
"""

from typing import Dict, Any, List, Tuple
from enum import Enum

from app.core.settings_store import get_runtime_pipeline_config


class MatchDecision(str, Enum):
    ACCEPTED = "accepted"  # Score >= accept_threshold (default 0.8)
    REVIEW = "to_review"    # review_threshold <= Score < accept_threshold
    REJECTED = "rejected"   # Score < review_threshold (default 0.5)


def compute_match_score(
    cv_skills: List[str],
    job_skills: List[str],
    cv_years: int = 0,
    job_years: int = 0,
    cv_edu_level: int = 2,
    job_edu_level: int = 2,
    similarity_score: float = 0.0,  # from semantic matching 0..1
) -> float:
    """Compute calibrated match score [0..1].

    Weights:
    - Skill matching: 50%
    - Semantic similarity: 20%
    - Experience: 15%
    - Education: 10%
    - Bonus: 5% for perfect match
    """
    cfg = get_runtime_pipeline_config()
    w_skill = cfg["skill_weight"]
    w_semantic = cfg["semantic_weight"]
    w_exp = cfg["experience_weight"]
    w_edu = cfg["education_weight"]
    bonus = cfg["perfect_match_bonus"]

    score = 0.0

    # Skill matching
    if job_skills:
        required_set = set(s.lower() for s in job_skills)
        cv_set = set(s.lower() for s in cv_skills)
        intersection = required_set & cv_set
        skill_score = len(intersection) / len(required_set) if intersection else 0.0
        score += skill_score * w_skill
    else:
        score += w_skill  # no skills required

    # Semantic similarity
    score += (similarity_score or 0.0) * w_semantic

    # Experience match
    if job_years > 0:
        if cv_years >= job_years:
            score += w_exp
        else:
            # Linear penalty: each missing year reduces the experience weight
            penalty = (job_years - cv_years) / job_years
            score += max(0, w_exp * (1 - penalty))
    else:
        score += w_exp

    # Education match
    if job_edu_level > 0:
        if cv_edu_level >= job_edu_level:
            score += w_edu
        else:
            # Penalty proportional to gap
            penalty = (job_edu_level - cv_edu_level) / job_edu_level
            score += max(0, w_edu * (1 - penalty))
    else:
        score += w_edu

    # Bonus for perfect skill + experience match
    if job_skills and cv_years >= job_years and len(intersection) == len(required_set):
        score += bonus

    return min(1.0, max(0.0, score))


def decide_match(score: float) -> MatchDecision:
    """Map score to decision using the configured thresholds."""
    cfg = get_runtime_pipeline_config()
    if score >= cfg["accept_threshold"]:
        return MatchDecision.ACCEPTED
    elif score >= cfg["review_threshold"]:
        return MatchDecision.REVIEW
    else:
        return MatchDecision.REJECTED


def get_decision_explanation(
    decision: MatchDecision,
    score: float,
    skill_match: float,
    experience_gap: int,
    missing_skills: List[str],
) -> str:
    """Generate human-readable explanation for the decision."""
    if decision == MatchDecision.ACCEPTED:
        msg = f"✅ Strong match (score: {score:.1%}). Candidate has the required experience and skills."
    elif decision == MatchDecision.REVIEW:
        msg = f"🟠 Worth reviewing (score: {score:.1%}). Some experience or skill gaps but overall promising."
    else:
        msg = f"❌ Not a match (score: {score:.1%}). Significant gaps in skills or experience."

    if missing_skills:
        msg += f"\n⚠️  Missing skills: {', '.join(missing_skills[:3])}"
    if experience_gap > 0:
        msg += f"\n📅 Experience gap: {experience_gap} years below requirement"

    return msg


def apply_business_rules(match_info: Dict[str, Any]) -> Dict[str, Any]:
    """Apply calibrated business rules and return enriched decision.

    Expected input keys:
    - score: float [0..1]
    - cv_skills: List[str]
    - job_skills: List[str]
    - cv_years: int
    - job_years: int
    - cv_edu: int (0-4)
    - job_edu: int (0-4)
    """
    score = match_info.get("score", 0.0)
    decision = decide_match(score)

    skill_match = 0.0
    missing = []
    if match_info.get("job_skills"):
        req = set(s.lower() for s in match_info.get("job_skills", []))
        cv = set(s.lower() for s in match_info.get("cv_skills", []))
        intersection = req & cv
        skill_match = len(intersection) / len(req) if req else 0
        missing = list(req - cv)

    exp_gap = max(0, match_info.get("job_years", 0) - match_info.get("cv_years", 0))

    explanation = get_decision_explanation(
        decision=decision,
        score=score,
        skill_match=skill_match,
        experience_gap=exp_gap,
        missing_skills=missing,
    )

    return {
        "decision": decision.value,
        "score": score,
        "skill_match_ratio": skill_match,
        "experience_gap_years": exp_gap,
        "missing_skills": missing,
        "explanation": explanation,
    }


__all__ = [
    "MatchDecision",
    "compute_match_score",
    "decide_match",
    "get_decision_explanation",
    "apply_business_rules",
]


if __name__ == "__main__":
    # Quick test
    result = apply_business_rules({
        "score": 0.75,
        "cv_skills": ["React", "Python", "AWS", "Docker"],
        "job_skills": ["React", "Node.js", "AWS"],
        "cv_years": 5,
        "job_years": 3,
        "cv_edu": 2,
        "job_edu": 2,
    })
    print(result)