import "server-only";

import type { NormalizedStoryboardScript, NormalizedStoryboardShot } from "@/lib/video-agent/types";

function detectLanguage(text: string): "zh" | "en" | "mixed" {
  const source = String(text || "");
  const zhCount = (source.match(/[\u4e00-\u9fff]/g) || []).length;
  const enCount = (source.match(/[A-Za-z]/g) || []).length;
  if (zhCount > 0 && enCount > 0) return "mixed";
  if (zhCount > 0) return "zh";
  return "en";
}

function cleanLine(line: string) {
  return String(line || "")
    .replace(/^\s*(?:第?\s*\d+\s*(?:镜头|段|幕)|\d+\s*[.)、：:\-]?|[一二三四五六七八九十]+\s*[、.：:\-]?)\s*/g, "")
    .replace(/\s+/g, " ")
    .trim();
}

function splitScriptLines(script: string): string[] {
  const source = String(script || "").trim();
  if (!source) return [];
  const byLine = source
    .split(/\r?\n+/)
    .map(cleanLine)
    .filter(Boolean);
  if (byLine.length >= 4) return byLine;

  return source
    .split(/[。！？!?；;\n]+/g)
    .map(cleanLine)
    .filter(Boolean);
}

function parsePreferredTileIndex(text: string, maxShots: number): number | undefined {
  const source = String(text || "");
  const patterns = [
    /第\s*(\d{1,2})\s*(?:张|格|图|镜头)/i,
    /(?:图|格|镜头|shot)\s*(\d{1,2})/i,
    /#\s*(\d{1,2})/i,
  ];
  for (const pattern of patterns) {
    const matched = source.match(pattern);
    const candidate = Number(matched?.[1] || 0);
    if (candidate >= 1 && candidate <= maxShots) return candidate;
  }
  return undefined;
}

function inferCamera(text: string) {
  const source = String(text || "");
  if (/推进|推近|close[- ]?in|dolly in/i.test(source)) return "轻微推进";
  if (/拉远|远景|zoom out|wide/i.test(source)) return "缓慢拉远";
  if (/平移|横移|pan/i.test(source)) return "平稳平移";
  if (/旋转|环绕|orbit/i.test(source)) return "环绕镜头";
  return "稳定镜头";
}

function inferMood(text: string) {
  const source = String(text || "");
  if (/紧张|压迫|惊悚|tense/i.test(source)) return "紧张";
  if (/温暖|治愈|暖/i.test(source)) return "温暖";
  if (/活力|动感|快节奏|energetic/i.test(source)) return "活力";
  if (/浪漫|梦幻|romantic/i.test(source)) return "浪漫";
  return "自然";
}

function buildDefaultAction(index: number) {
  const defaults = [
    "主体从静止到轻微动作",
    "主体与环境产生互动",
    "主体视线发生转移",
    "主体做细节动作展示",
    "镜头展示关键部位",
    "镜头拉近突出质感",
    "镜头切换到侧面动态",
    "主体完成动作收束",
    "镜头定格形成结尾",
  ];
  return defaults[index - 1] || `主体完成第 ${index} 段自然动作`;
}

export function runScriptNormalizerSkill(params: {
  scriptText: string;
  maxShots?: number;
  totalDurationSec?: number;
}): NormalizedStoryboardScript {
  const maxShots = Math.max(1, Math.min(20, Math.round(params.maxShots || 9)));
  const raw = String(params.scriptText || "").trim();
  const lines = splitScriptLines(raw);
  const effectiveLines = lines.slice(0, maxShots);
  const shotCount = Math.max(1, effectiveLines.length);
  const durationPerShot = Math.max(
    2,
    Math.round(Math.max(5, Number(params.totalDurationSec || 10)) / shotCount)
  );

  const normalizedShots: NormalizedStoryboardShot[] = [];
  const language = detectLanguage(raw);
  const fallbackStyle =
    language === "en" ? "cinematic realistic style" : "写实电影感";

  for (let index = 1; index <= shotCount; index += 1) {
    const line = effectiveLines[index - 1] || "";
    const action = cleanLine(line) || buildDefaultAction(index);
    normalizedShots.push({
      index,
      durationSec: durationPerShot,
      action,
      camera: inferCamera(line),
      mood: inferMood(line),
      negativePrompt:
        language === "en"
          ? "jitter, blur, artifacts, low quality"
          : "抖动, 畸变, 低清, 伪影",
      preferredTileIndex: parsePreferredTileIndex(line, maxShots),
    });
  }

  return {
    language,
    globalStyle: fallbackStyle,
    shotCount: normalizedShots.length,
    shots: normalizedShots,
    raw,
  };
}