"""Hiring-fit prediction service."""

from app.schemas.common import JobStatus
from app.schemas.hiring_fit import HiringFitRequest, HiringFitResponse
from app.services.ai_generate import generate_json


def _label_and_pass(score: float) -> tuple[str, bool]:
    if score >= 0.85:
        return "strong_yes", True
    if score >= 0.70:
        return "yes", True
    if score >= 0.50:
        return "maybe", False
    return "no", False


class HiringFitService:
    async def predict(self, payload: HiringFitRequest) -> HiringFitResponse:
        system = (
            "Predict hiring fit. Return JSON: {\"hiring_fit_score\":0-1,"
            "\"hiring_fit_label\",\"reasons\":[],\"risks\":[],\"next_step\"}"
        )
        user = (
            f"Candidate: {payload.candidate_id}\nRole: {payload.role_title}\n"
            f"JD: {payload.job_description}\nSkills: {payload.required_skills}\n"
            f"Resume: {payload.resume_text}\nScorecard: {payload.scorecard}\n"
            f"Transcript turns: {len(payload.transcript)}"
        )

        def local() -> dict:
            base = 0.6
            if payload.scorecard:
                vals = [float(v) for v in payload.scorecard.values() if isinstance(v, (int, float))]
                if vals:
                    base = sum(vals) / len(vals)
            if payload.required_skills and payload.resume_text:
                hits = sum(1 for s in payload.required_skills if s.lower() in payload.resume_text.lower())
                base = min(0.95, base + 0.05 * hits)
            label, _ = _label_and_pass(base)
            return {
                "hiring_fit_score": round(base, 2),
                "hiring_fit_label": label,
                "reasons": ["Profile partially aligns with role requirements"],
                "risks": ["Validate depth with a focused technical follow-up"],
                "next_step": "advance" if base >= 0.70 else "hold_for_follow_up",
            }

        data, stub, provider = await generate_json(
            system=system, user=user, local_factory=local
        )
        score = float(data.get("hiring_fit_score") or 0.5)
        label = data.get("hiring_fit_label") or _label_and_pass(score)[0]
        passed = score >= 0.70
        return HiringFitResponse(
            correlation_id=payload.correlation_id,
            candidate_id=payload.candidate_id,
            role_title=payload.role_title,
            status=JobStatus.SUCCEEDED,
            stub=stub,
            message=f"Hiring-fit prediction via {provider}",
            hiring_fit_score=score,
            hiring_fit_label=label,
            passed=passed,
            reasons=list(data.get("reasons") or []),
            risks=list(data.get("risks") or []),
            next_step=data.get("next_step"),
        )