c/backend/app/routes/spam_routes.py

from flask import Blueprint, current_app, request
from flask_jwt_extended import jwt_required

from app.extensions import db
from app.ml.naive_bayes_classifier import NaiveBayesSpamClassifier
from app.ml.spam_categorizer import categorize_spam, get_category_label
from app.models import DetectionConfig, SpamPredictionLog, SpamTrainingSample
from app.utils.auth import admin_required, current_user
from app.utils.response import fail, ok


spam_bp = Blueprint("spam", __name__)


def _classifier() -> NaiveBayesSpamClassifier:
    return NaiveBayesSpamClassifier(current_app.config["NB_MODEL_PATH"])


def _active_samples() -> list[dict]:
    rows = SpamTrainingSample.query.filter_by(is_active=True).order_by(SpamTrainingSample.id.asc()).all()
    return [{"text": row.text, "label": row.label} for row in rows]


def _ensure_ready() -> NaiveBayesSpamClassifier:
    clf = _classifier()
    samples = _active_samples()
    clf.ensure_ready(samples)
    return clf


def _threshold() -> float:
    row = DetectionConfig.query.order_by(DetectionConfig.id.asc()).first()
    return float(row.spam_threshold) if row else 0.75


def _adjusted_threshold(user_credit: int = 100) -> float:
    """根据用户信誉分调整阈值"""
    base_threshold = _threshold()
    # 系数范围：0.85 - 1.15
    credit_factor = 1.0 + (user_credit - 100) * 0.0015
    credit_factor = max(0.85, min(1.15, credit_factor))
    return base_threshold * credit_factor


@spam_bp.post("/predict")
@jwt_required()
def predict_one():
    user = current_user()
    if not user:
        return fail("用户不存在", 404)

    payload = request.get_json(silent=True) or {}
    text = (payload.get("text") or "").strip()
    if len(text) < 2:
        return fail("请输入至少2个字符的待识别文本", 400)

    clf = _ensure_ready()
    result = clf.predict(text)
    threshold = _adjusted_threshold(user.credit_score or 100)
    blocked = float(result["spam_probability"]) >= threshold

    # 分类标签：仅在判定为垃圾时进行细分
    category = ""
    category_label = ""
    if blocked:
        category, category_label = categorize_spam(result["text"])

    row = SpamPredictionLog(
        user_id=user.id,
        text=result["text"],
        prediction=result["prediction"],
        category=category,
        spam_probability=result["spam_probability"],
        ham_probability=result["ham_probability"],
        confidence=result["confidence"],
        reason_tokens=result["reason_tokens"],
        model_version=result.get("model_version", ""),
    )
    db.session.add(row)
    db.session.commit()

    return ok({
        **result,
        "log_id": row.id,
        "threshold": threshold,
        "blocked_by_threshold": blocked,
        "category": category,
        "category_label": category_label,
    }, "识别成功")


@spam_bp.post("/predict/batch")
@jwt_required()
def predict_batch():
    user = current_user()
    if not user:
        return fail("用户不存在", 404)

    payload = request.get_json(silent=True) or {}
    items = payload.get("items") or []
    if not isinstance(items, list) or not items:
        return fail("items 必须是非空数组", 400)
    if len(items) > 100:
        return fail("单次最多识别100条", 400)

    clf = _ensure_ready()
    rows = []
    results = []
    threshold = _adjusted_threshold(user.credit_score or 100)

    for text in items:
        content = (text or "").strip()
        if len(content) < 2:
            continue
        result = clf.predict(content)
        blocked = float(result["spam_probability"]) >= threshold
        result["blocked_by_threshold"] = blocked

        # 分类标签
        category = ""
        category_label = ""
        if blocked:
            category, category_label = categorize_spam(result["text"])
        result["category"] = category
        result["category_label"] = category_label

        rows.append(
            SpamPredictionLog(
                user_id=user.id,
                text=result["text"],
                prediction=result["prediction"],
                category=category,
                spam_probability=result["spam_probability"],
                ham_probability=result["ham_probability"],
                confidence=result["confidence"],
                reason_tokens=result["reason_tokens"],
                model_version=result.get("model_version", ""),
            )
        )
        results.append(result)

    if not rows:
        return fail("没有可识别的有效文本", 400)

    db.session.add_all(rows)
    db.session.commit()

    spam_count = len([item for item in results if item["prediction"] == "spam"])
    blocked_count = len([item for item in results if item["blocked_by_threshold"]])
    ham_count = len(results) - spam_count

    return ok(
        {
            "items": results,
            "summary": {
                "total": len(results),
                "spam_count": spam_count,
                "ham_count": ham_count,
                "blocked_count": blocked_count,
                "spam_ratio": round(spam_count / len(results), 4) if results else 0,
                "blocked_ratio": round(blocked_count / len(results), 4) if results else 0,
                "threshold": threshold,
            },
        },
        "批量识别完成",
    )


@spam_bp.get("/history")
@jwt_required()
def my_history():
    user = current_user()
    if not user:
        return fail("用户不存在", 404)

    page = max(int(request.args.get("page", 1) or 1), 1)
    page_size = min(max(int(request.args.get("page_size", 20) or 20), 1), 100)

    pagination = (
        SpamPredictionLog.query.filter_by(user_id=user.id)
        .order_by(SpamPredictionLog.id.desc())
        .paginate(page=page, per_page=page_size, error_out=False)
    )

    return ok(
        {
            "items": [item.to_dict() for item in pagination.items],
            "total": pagination.total,
            "page": page,
            "page_size": page_size,
        }
    )


@spam_bp.delete("/history/<int:log_id>")
@jwt_required()
def delete_history(log_id: int):
    user = current_user()
    if not user:
        return fail("用户不存在", 404)

    row = SpamPredictionLog.query.filter_by(id=log_id, user_id=user.id).first()
    if not row:
        return fail("记录不存在", 404)

    db.session.delete(row)
    db.session.commit()
    return ok({}, "记录已删除")


@spam_bp.post("/feedback")
@jwt_required()
def save_feedback():
    user = current_user()
    if not user:
        return fail("用户不存在", 404)

    payload = request.get_json(silent=True) or {}
    text = (payload.get("text") or "").strip()
    label = NaiveBayesSpamClassifier.normalize_label(payload.get("label"))

    if len(text) < 2:
        return fail("文本至少2个字符", 400)
    if not label:
        return fail("label 必须是 spam 或 ham", 400)

    row = SpamTrainingSample(text=text, label=label, source="feedback", created_by=user.id, is_active=True)
    db.session.add(row)
    db.session.commit()
    return ok(row.to_dict(), "反馈样本已记录")


@spam_bp.get("/model/info")
@jwt_required(optional=True)
def model_info():
    clf = _classifier()
    clf.load()
    info = clf.model_info()
    info["threshold"] = _threshold()
    return ok(info)


@spam_bp.post("/train")
@admin_required
def train_model():
    clf = _classifier()
    samples = _active_samples()
    metadata = clf.train(samples)
    return ok(metadata, "模型训练完成")


@spam_bp.get("/samples")
@admin_required
def list_samples():
    keyword = (request.args.get("keyword") or "").strip()
    label = (request.args.get("label") or "").strip().lower()
    page = max(int(request.args.get("page", 1) or 1), 1)
    page_size = min(max(int(request.args.get("page_size", 20) or 20), 1), 100)

    query = SpamTrainingSample.query
    if keyword:
        query = query.filter(SpamTrainingSample.text.like(f"%{keyword}%"))
    if label in {"spam", "ham"}:
        query = query.filter(SpamTrainingSample.label == label)

    pagination = query.order_by(SpamTrainingSample.id.desc()).paginate(page=page, per_page=page_size, error_out=False)
    return ok(
        {
            "items": [item.to_dict() for item in pagination.items],
            "total": pagination.total,
            "page": page,
            "page_size": page_size,
        }
    )


@spam_bp.post("/samples")
@admin_required
def create_sample():
    payload = request.get_json(silent=True) or {}
    text = (payload.get("text") or "").strip()
    label = NaiveBayesSpamClassifier.normalize_label(payload.get("label"))

    if len(text) < 2:
        return fail("文本至少2个字符", 400)
    if not label:
        return fail("label 必须是 spam 或 ham", 400)

    user = current_user()
    row = SpamTrainingSample(text=text, label=label, source="import", created_by=user.id if user else None, is_active=True)
    db.session.add(row)
    db.session.commit()
    return ok(row.to_dict(), "样本创建成功")


@spam_bp.put("/samples/<int:sample_id>")
@admin_required
def update_sample(sample_id: int):
    row = SpamTrainingSample.query.get(sample_id)
    if not row:
        return fail("样本不存在", 404)

    payload = request.get_json(silent=True) or {}
    if "text" in payload:
        text = (payload.get("text") or "").strip()
        if len(text) < 2:
            return fail("文本至少2个字符", 400)
        row.text = text
    if "label" in payload:
        label = NaiveBayesSpamClassifier.normalize_label(payload.get("label"))
        if not label:
            return fail("label 必须是 spam 或 ham", 400)
        row.label = label
    if "is_active" in payload:
        row.is_active = bool(payload.get("is_active"))

    db.session.commit()
    return ok(row.to_dict(), "样本更新成功")


@spam_bp.delete("/samples/<int:sample_id>")
@admin_required
def delete_sample(sample_id: int):
    row = SpamTrainingSample.query.get(sample_id)
    if not row:
        return fail("样本不存在", 404)

    db.session.delete(row)
    db.session.commit()
    return ok({}, "样本已删除")


@spam_bp.post("/samples/import")
@admin_required
def import_samples():
    payload = request.get_json(silent=True) or {}
    items = payload.get("items") or []
    if not isinstance(items, list) or not items:
        return fail("items 必须是非空数组", 400)

    user = current_user()
    created = 0
    updated = 0

    for item in items:
        text = (item.get("text") or "").strip()
        label = NaiveBayesSpamClassifier.normalize_label(item.get("label"))
        if len(text) < 2 or not label:
            continue

        row = SpamTrainingSample.query.filter_by(text=text).first()
        if row:
            row.label = label
            row.is_active = bool(item.get("is_active", True))
            row.source = item.get("source") or row.source
            updated += 1
        else:
            row = SpamTrainingSample(
                text=text,
                label=label,
                source=item.get("source") or "import",
                created_by=user.id if user else None,
                is_active=bool(item.get("is_active", True)),
            )
            db.session.add(row)
            created += 1

    db.session.commit()
    return ok({"created": created, "updated": updated}, "样本导入完成")