ai-chatbot/app/(chat)/api/chat/_lib/video-models.ts

import { createLogger } from "@/lib/observability/logger";
import { after } from "next/server";
import {
  getAiModelById,
  getEnabledAiModelsWithProvider,
  updateMessageById,
} from "@/lib/db/queries";
import {
  isImageGenerationModelDescriptor,
  isVideoModelDescriptor,
} from "@/lib/ai/model-classification";
import type { StoryboardModelCandidate } from "@/lib/video-agent/types";

const chatApiLogger = createLogger("chat.videoModels");


export type VideoModeType =
  | "text-to-video"
  | "first-frame"
  | "first-last-frame"
  | "reference-images";

export function isVideoProviderType(
  providerType?: string | null,
  model?: {
    modelId?: string | null;
    capabilities?: unknown;
    supportsImageGen?: boolean | null;
  },
): boolean {
  return isVideoModelDescriptor({
    providerType,
    modelId: model?.modelId,
    capabilities: model?.capabilities,
    supportsImageGen: model?.supportsImageGen,
  });
}

export function isImageGenerationModelType(
  providerType?: string | null,
  model?: {
    modelId?: string | null;
    capabilities?: unknown;
    supportsImageGen?: boolean | null;
  },
): boolean {
  return isImageGenerationModelDescriptor({
    providerType,
    modelId: model?.modelId,
    capabilities: model?.capabilities,
    supportsImageGen: model?.supportsImageGen,
  });
}

export function inferSupportedVideoModesByModelId(modelId: string): VideoModeType[] {
  const normalized = modelId.toLowerCase();

  if (
    normalized === "gen4.5" ||
    normalized === "gen4_turbo" ||
    normalized === "gen4_aleph" ||
    normalized === "gen3a_turbo" ||
    normalized === "veo3" ||
    normalized === "veo3.1" ||
    normalized === "veo3.1_fast"
  ) {
    return ["text-to-video", "first-frame"];
  }

  if (normalized.includes("lite-t2v")) {
    return ["text-to-video"];
  }
  if (normalized.includes("lite-i2v")) {
    return ["first-frame", "first-last-frame", "reference-images"];
  }
  if (normalized.includes("pro-fast")) {
    return ["text-to-video", "first-frame"];
  }
  if (normalized.includes("-pro-") || normalized.endsWith("-pro")) {
    return ["text-to-video", "first-frame", "first-last-frame"];
  }

  return ["text-to-video", "first-frame", "first-last-frame", "reference-images"];
}

export function getSupportedVideoModesForModel(model: any): VideoModeType[] {
  const capabilityModes = (((model?.capabilities as any)?.supportedModes ?? []) as string[]).filter(
    (mode): mode is VideoModeType =>
      mode === "text-to-video" ||
      mode === "first-frame" ||
      mode === "first-last-frame" ||
      mode === "reference-images",
  );

  if (capabilityModes.length > 0) {
    return capabilityModes;
  }

  const modelId = String(model?.modelId || "");
  return inferSupportedVideoModesByModelId(modelId);
}

export function scoreVideoSubModelForMode(modelId: string, mode: VideoModeType): number {
  const normalized = modelId.toLowerCase();
  const isProFast = normalized.includes("pro-fast");
  const isLiteT2V = normalized.includes("lite-t2v");
  const isLiteI2V = normalized.includes("lite-i2v");
  const isPro = (normalized.includes("-pro-") || normalized.endsWith("-pro")) && !isProFast;
  const isV15Pro = normalized.includes("seedance-1-5-pro");

  switch (mode) {
    case "first-last-frame":
      if (isV15Pro) return 150;
      if (isPro) return 120;
      if (isLiteI2V) return 90;
      if (isProFast) return -1000;
      if (isLiteT2V) return -1000;
      return 10;
    case "first-frame":
      if (isV15Pro) return 140;
      if (isProFast) return 130;
      if (isPro) return 120;
      if (isLiteI2V) return 100;
      if (isLiteT2V) return 20;
      return 10;
    case "reference-images":
      if (isV15Pro) return 150;
      if (isLiteI2V) return 140;
      if (isPro) return 110;
      if (isProFast || isLiteT2V) return -1000;
      return 10;
    case "text-to-video":
    default:
      if (isV15Pro) return 140;
      if (isLiteT2V) return 130;
      if (isProFast) return 120;
      if (isPro) return 100;
      if (isLiteI2V) return 80;
      return 10;
  }
}

export function selectBestVideoSubModelForMode(
  subModels: Array<{ model: any }>,
  mode: VideoModeType,
): { model: any } | null {
  const compatible = subModels.filter(({ model }) => {
    const supportedModes = getSupportedVideoModesForModel(model);
    return supportedModes.includes(mode);
  });

  const candidates = compatible.length > 0 ? compatible : subModels;
  if (candidates.length === 0) return null;

  return [...candidates].sort((a, b) => {
    const scoreA = scoreVideoSubModelForMode(String(a.model?.modelId || ""), mode);
    const scoreB = scoreVideoSubModelForMode(String(b.model?.modelId || ""), mode);
    return scoreB - scoreA;
  })[0];
}

export async function resolveStoryboardModelCandidates(
  selectedModelRef: string,
): Promise<StoryboardModelCandidate[]> {
  const selectedModel = await getAiModelById({ id: selectedModelRef }).catch(() => undefined);
  if (!selectedModel) return [];

  const configuredFallbackIds = [
    selectedModel.fallbackModelId1,
    selectedModel.fallbackModelId2,
  ].filter((item): item is string => Boolean(String(item || "").trim()));

  const allModels = await getEnabledAiModelsWithProvider(true, true).catch(() => []);
  const sameProvider = allModels
    .filter(
      ({ model, provider }: any) =>
        provider?.id === selectedModel.providerId && model?.id !== selectedModel.id && model?.isEnabled,
    )
    .map(({ model }: any) => String(model?.id || "").trim())
    .filter(Boolean);

  const candidateIds = Array.from(new Set([selectedModel.id, ...configuredFallbackIds, ...sameProvider]));
  const byModelRef = new Map<string, any>();
  allModels
    .map(({ model }: any) => model)
    .filter(Boolean)
    .forEach((model: any) => {
      byModelRef.set(String(model.id), model);
      if (model.modelId) {
        byModelRef.set(String(model.modelId), model);
      }
    });

  return candidateIds
    .map((id) => byModelRef.get(id))
    .filter(Boolean)
    .map((model: any) => ({
      id: String(model.id),
      modelId: String(model.modelId || "").trim(),
      displayName: String(model.displayName || model.modelId || "").trim(),
    }))
    .filter((item) => item.modelId);
}

export function extractVideoUrlFromText(text: string): string | null {
  const normalizedText = String(text || "");
  if (!normalizedText) return null;

  const htmlMatch = normalizedText.match(/<video[^>]+src="([^"]+)"/i);
  if (htmlMatch?.[1]) return htmlMatch[1];

  const markdownMatch = normalizedText.match(
    /\[[^\]]*]\((https?:\/\/[^)\s]+\.mp4(?:\?[^)\s]*)?)\)/i,
  );
  if (markdownMatch?.[1]) return markdownMatch[1];

  const plainMatch = normalizedText.match(/(https?:\/\/[^\s"'<>]+\.mp4(?:\?[^\s"'<>]+)?)/i);
  return plainMatch?.[1] || null;
}

export function isAlreadyArchivedVideoUrl(url: string): boolean {
  return /\/uploads\//i.test(url);
}

export function scheduleVideoArchive(params: {
  chatId: string;
  messageId: string;
  text: string;
  temporaryVideoUrl: string;
  taskId?: string;
  metadata?: Record<string, any> | null;
}) {
  const { chatId, messageId, temporaryVideoUrl, taskId, metadata } = params;
  if (!temporaryVideoUrl || isAlreadyArchivedVideoUrl(temporaryVideoUrl)) {
    return;
  }

  const delayHoursRaw = Number(process.env.VIDEO_ARCHIVE_DELAY_HOURS || 48);
  const delayHours = Number.isFinite(delayHoursRaw)
    ? Math.max(1, Math.min(24 * 30, Math.round(delayHoursRaw)))
    : 48;
  const archiveAfterAtMs = Date.now() + delayHours * 60 * 60 * 1000;

  after(async () => {
    try {
      const nextMetadata = {
        ...(metadata || {}),
        temporaryVideoUrl,
        storageStatus: "temporary",
        archiveMode: "delayed_job",
        archiveAfterAt: archiveAfterAtMs,
        archiveDelayHours: delayHours,
        archiveTaskId: taskId || messageId,
        updatedAt: Date.now(),
      };

      await updateMessageById({
        messageId,
        chatId,
        videoStorageStatus: "temporary",
        temporaryVideoUrl,
        permanentVideoUrl: null,
        archivedAt: null,
        videoMetadata: nextMetadata as any,
      });

      chatApiLogger.info("[Chat API] 🕒 已登记延迟归档视频:", {
        messageId,
        archiveAfterAt: new Date(archiveAfterAtMs).toISOString(),
      });
    } catch (error) {
      chatApiLogger.error("[Chat API] ❌ 登记视频延迟归档失败（保留临时地址）:", {
        messageId,
        temporaryVideoUrl,
        error,
      });
    }
  });
}