ai-chatbot/lib/video-agent/storyboard-command-interpreter.ts

import "server-only";

import { generateText } from "ai";
import { getLanguageModelById } from "@/lib/ai/dynamic-provider";
import { getAgentAiModelWithProvider, getEnabledAiModelsWithProvider } from "@/lib/db/queries";

export type StoryboardCommandIntent = "confirm" | "revise" | "cancel" | "unclear";

export type StoryboardRevisionCommand =
  | {
      type: "update_prompt";
      shotIndex: number;
      prompt: string;
    }
  | {
      type: "regenerate_hd";
      shotIndex: number;
    }
  | null;

export interface StoryboardCommandContextItem {
  index: number;
  title?: string;
  summary?: string;
}

const CONFIRM_PATTERNS = [
  /确认开始生成视频/,
  /确认开始生成/,
  /按这个分镜生成/,
  /就按这个生成/,
  /按这个来生成/,
  /开始生成视频/,
  /开始吧/,
  /可以开始了/,
  /可以执行了/,
  /直接生成吧/,
  /直接开始吧/,
  /先这样/,
  /没问题开始/,
  /确认无误/,
  /就这样吧/,
];

const CANCEL_PATTERNS = [
  /取消/,
  /先不要/,
  /暂停/,
  /停止/,
  /先别生成/,
  /先别做/,
  /不用生成了/,
  /不用做了/,
];

const REVISE_HINT_PATTERNS = [
  /修改/,
  /调整/,
  /改成/,
  /改为/,
  /换成/,
  /换成/,
  /做成/,
  /重生/,
  /重做/,
  /重新生成/,
  /重新做/,
  /不要/,
  /太糊/,
  /太平/,
  /太死板/,
];

function normalizeCompactText(text: string) {
  return String(text || "").toLowerCase().replace(/\s+/g, "").trim();
}

function chineseNumberToInt(source: string): number | null {
  const normalized = String(source || "").trim();
  if (!normalized) return null;
  if (/^\d+$/.test(normalized)) {
    const parsed = Number(normalized);
    return Number.isFinite(parsed) ? parsed : null;
  }

  const map: Record<string, number> = {
    一: 1,
    二: 2,
    两: 2,
    三: 3,
    四: 4,
    五: 5,
    六: 6,
    七: 7,
    八: 8,
    九: 9,
    十: 10,
  };

  if (normalized === "十") return 10;
  if (normalized.length === 1) {
    return map[normalized] ?? null;
  }
  if (/^十[一二三四五六七八九]$/.test(normalized)) {
    return 10 + (map[normalized.slice(1)] || 0);
  }
  if (/^[一二三四五六七八九]十$/.test(normalized)) {
    return (map[normalized[0]!] || 0) * 10;
  }
  if (/^[一二三四五六七八九]十[一二三四五六七八九]$/.test(normalized)) {
    return (map[normalized[0]!] || 0) * 10 + (map[normalized[2]!] || 0);
  }

  return null;
}

function resolveShotIndexFromText(text: string, totalShots?: number): number | null {
  const normalized = String(text || "").trim();
  if (!normalized) return null;

  const explicitPatterns = [
    /第\s*([一二三四五六七八九十两\d]{1,3})\s*(?:张|格|图|镜头|段)/i,
    /把\s*([一二三四五六七八九十两\d]{1,3})\s*(?:张|格|图|镜头|段)/i,
    /([一二三四五六七八九十两\d]{1,3})\s*(?:号)?(?:张|格|图|镜头|段)/i,
  ];

  for (const pattern of explicitPatterns) {
    const match = normalized.match(pattern);
    if (!match?.[1]) continue;
    const resolved = chineseNumberToInt(match[1]);
    if (resolved && resolved >= 1 && (!totalShots || resolved <= totalShots)) {
      return resolved;
    }
  }

  if (totalShots && /最后一(?:张|格|图|镜头|段)/.test(normalized)) {
    return totalShots;
  }

  if (totalShots) {
    const reverseMatch = normalized.match(/倒数第\s*([一二三四五六七八九十两\d]{1,3})\s*(?:张|格|图|镜头|段)/i);
    if (reverseMatch?.[1]) {
      const reversed = chineseNumberToInt(reverseMatch[1]);
      if (reversed && reversed >= 1 && reversed <= totalShots) {
        return totalShots - reversed + 1;
      }
    }
  }

  return null;
}

function extractRevisionPrompt(text: string): string {
  return String(text || "")
    .replace(/^(?:请)?(?:把)?\s*第?\s*[一二三四五六七八九十两\d]{1,3}\s*(?:张|格|图|镜头|段)(?:动作)?\s*(?:修改|调整|改成|改为|换成|做成)?\s*[:：]?\s*/i, "")
    .replace(/^(?:请)?(?:把)?\s*最后一(?:张|格|图|镜头|段)\s*(?:修改|调整|改成|改为|换成|做成)?\s*[:：]?\s*/i, "")
    .trim();
}

export function detectStoryboardCommandIntent(text: string): StoryboardCommandIntent {
  const normalized = normalizeCompactText(text);
  if (!normalized) return "unclear";
  if (CANCEL_PATTERNS.some((pattern) => pattern.test(normalized))) return "cancel";
  if (CONFIRM_PATTERNS.some((pattern) => pattern.test(normalized))) return "confirm";
  if (REVISE_HINT_PATTERNS.some((pattern) => pattern.test(normalized))) return "revise";
  if (resolveShotIndexFromText(normalized) !== null) return "revise";
  return "unclear";
}

function parseJsonBlock<T = Record<string, unknown>>(text: string): T | null {
  const match = String(text || "").match(/\{[\s\S]*\}/);
  if (!match) return null;
  try {
    return JSON.parse(match[0]) as T;
  } catch {
    return null;
  }
}

type EnabledModelWithProvider = Awaited<
  ReturnType<typeof import("@/lib/db/queries")["getEnabledAiModelsWithProvider"]>
>[number];

function isAgentGenerateTextProviderType(providerType: string): boolean {
  return ["zhipu", "alibaba", "openai", "custom", "local"].includes(providerType);
}

function isAgentGenerateTextCompatibleModel(item: EnabledModelWithProvider) {
  if (!item?.model?.isEnabled || item.model.isSubModel) return false;
  return isAgentGenerateTextProviderType(String(item.provider?.type || ""));
}

function pickReasoningModel(
  configuredAgentModel: EnabledModelWithProvider | undefined,
  models: EnabledModelWithProvider[],
) {
  if (
    configuredAgentModel &&
    isAgentGenerateTextCompatibleModel(configuredAgentModel)
  ) {
    return configuredAgentModel;
  }

  const candidates = models.filter((item) => isAgentGenerateTextCompatibleModel(item));
  const zhipuModel = candidates.find((item) => item.provider.type === "zhipu");
  return zhipuModel || candidates[0];
}

async function interpretStoryboardCommandWithLLM(params: {
  text: string;
  totalShots?: number;
  items?: StoryboardCommandContextItem[];
}): Promise<{ intent: StoryboardCommandIntent; revision: StoryboardRevisionCommand } | null> {
  try {
    const allModels = await getEnabledAiModelsWithProvider();
    const configuredAgentModel = await getAgentAiModelWithProvider();
    const reasoningModelConfig = pickReasoningModel(configuredAgentModel, allModels);
    if (!reasoningModelConfig) return null;

    const model = await getLanguageModelById(reasoningModelConfig.model.id);
    if (!model) return null;

    const contextLines = (params.items || [])
      .slice(0, 12)
      .map((item) => `- 镜头${item.index}：${item.summary || item.title || "无描述"}`)
      .join("\n");

    const { text } = await generateText({
      model,
      system: `你是一个分镜任务指令解释器。
你需要判断用户当前这句话，是在：
- confirm：确认开始执行
- revise：修改某个镜头或重生某张高清图
- cancel：取消当前流程
- unclear：无法判断

只输出 JSON，不要输出解释：
{
  "intent": "confirm|revise|cancel|unclear",
  "action": "update_prompt|regenerate_hd|null",
  "shotIndex": 1,
  "prompt": "修改后的镜头文案，没有就空字符串"
}

规则：
1. 用户表达“开始、确认、就这样、按这个生成”时，intent=confirm。
2. 用户表达“取消、暂停、先别生成”时，intent=cancel。
3. 用户表达“改一下、调整、换成、重生、重做、重新生成”且指向某镜头时，intent=revise。
4. 如果用户说“最后一张/最后一格/最后一个镜头”，shotIndex 取最后一个。
5. 如果用户说“倒数第二张/倒数第二个镜头”，按总镜头数推算。
6. 如果 intent=revise 且是文字修改，action=update_prompt。
7. 如果 intent=revise 且是重生高清预览，action=regenerate_hd。
8. 如果无法确定具体镜头，intent=unclear。`,
      prompt: [
        `总镜头数：${Math.max(0, Number(params.totalShots || 0))}`,
        contextLines ? `当前镜头草案：\n${contextLines}` : "",
        `用户输入：${params.text}`,
      ]
        .filter(Boolean)
        .join("\n\n"),
      temperature: 0.1,
      maxTokens: 200,
      abortSignal: AbortSignal.timeout(12000),
    } as any);

    const parsed = parseJsonBlock<{
      intent?: string;
      action?: string | null;
      shotIndex?: number;
      prompt?: string;
    }>(text);
    if (!parsed) return null;

    const intent =
      parsed.intent === "confirm" ||
      parsed.intent === "revise" ||
      parsed.intent === "cancel" ||
      parsed.intent === "unclear"
        ? parsed.intent
        : "unclear";

    if (intent !== "revise") {
      return { intent, revision: null };
    }

    const shotIndex = Math.max(1, Number(parsed.shotIndex || 0));
    if (!Number.isFinite(shotIndex) || shotIndex <= 0) {
      return { intent: "unclear", revision: null };
    }

    if (parsed.action === "regenerate_hd") {
      return {
        intent,
        revision: {
          type: "regenerate_hd",
          shotIndex,
        },
      };
    }

    const prompt = String(parsed.prompt || "").trim();
    if (!prompt) {
      return { intent: "unclear", revision: null };
    }

    return {
      intent,
      revision: {
        type: "update_prompt",
        shotIndex,
        prompt,
      },
    };
  } catch {
    return null;
  }
}

export function parseStoryboardRevisionCommand(params: {
  text: string;
  totalShots?: number;
}): StoryboardRevisionCommand {
  const source = String(params.text || "").trim();
  if (!source) return null;

  const shotIndex = resolveShotIndexFromText(source, params.totalShots);
  if (!shotIndex) return null;

  if (
    /重生|重做|重新生成|重新做|再生成一版|再来一版|重新出图|重出高清图|高清图重来/i.test(source)
  ) {
    return {
      type: "regenerate_hd",
      shotIndex,
    };
  }

  const prompt = extractRevisionPrompt(source);
  if (!prompt) {
    return null;
  }

  return {
    type: "update_prompt",
    shotIndex,
    prompt,
  };
}

export async function interpretStoryboardCommand(params: {
  text: string;
  totalShots?: number;
  items?: StoryboardCommandContextItem[];
}): Promise<{ intent: StoryboardCommandIntent; revision: StoryboardRevisionCommand }> {
  const heuristicIntent = detectStoryboardCommandIntent(params.text);
  const heuristicRevision = parseStoryboardRevisionCommand({
    text: params.text,
    totalShots: params.totalShots,
  });

  if (heuristicIntent === "confirm" || heuristicIntent === "cancel") {
    return { intent: heuristicIntent, revision: null };
  }

  if (heuristicIntent === "revise" && heuristicRevision) {
    return { intent: heuristicIntent, revision: heuristicRevision };
  }

  const llmInterpretation = await interpretStoryboardCommandWithLLM(params);
  if (llmInterpretation) {
    return llmInterpretation;
  }

  return {
    intent: heuristicIntent,
    revision: heuristicRevision,
  };
}