ai-chatbot/lib/openclaw-chat/media-attachments.ts

export const OPENCLAW_CHAT_VIDEO_UPLOAD_PURPOSE = "openclaw-chat-video";
export const OPENCLAW_CHAT_VIDEO_MAX_FILES = 2;
export const OPENCLAW_CHAT_VIDEO_MAX_BYTES = 500 * 1024 * 1024;
export const OPENCLAW_CHAT_MEDIA_MAX_ATTACHMENTS = 16;
export const OPENCLAW_CHAT_VIDEO_ACCEPT =
  "video/mp4,video/quicktime,video/webm,video/x-m4v,video/x-matroska,.mp4,.mov,.webm,.m4v,.mkv";

const VIDEO_CONTENT_TYPES = new Set([
  "video/mp4",
  "video/quicktime",
  "video/webm",
  "video/x-m4v",
  "video/x-matroska",
]);

const VIDEO_CONTENT_TYPE_BY_EXTENSION: Record<string, string> = {
  m4v: "video/x-m4v",
  mkv: "video/x-matroska",
  mov: "video/quicktime",
  mp4: "video/mp4",
  webm: "video/webm",
};

const VIDEO_EXTENSION_PATTERN = /\.(mp4|mov|webm|m4v|mkv)(?:$|[?#])/i;

export type OpenClawChatMediaAttachment = {
  type: "image" | "video";
  url: string;
  name?: string;
  content_type?: string;
};

type OpenClawChatAttachmentLike = {
  type?: unknown;
  url?: unknown;
  name?: unknown;
  contentType?: unknown;
  content_type?: unknown;
};

type OpenClawChatVideoFileLike = {
  name: string;
  type: string;
};

function normalizeContentType(value: unknown) {
  return String(value || "")
    .split(";", 1)[0]
    .trim()
    .toLowerCase();
}

function getExtension(value: string) {
  const cleanValue = String(value || "").split(/[?#]/, 1)[0];
  return cleanValue.split(".").pop()?.trim().toLowerCase() || "";
}

function isHttpUrl(value: string) {
  try {
    const parsed = new URL(value);
    return parsed.protocol === "http:" || parsed.protocol === "https:";
  } catch {
    return false;
  }
}

export function getOpenClawChatVideoContentType(
  file: OpenClawChatVideoFileLike
) {
  const contentType = normalizeContentType(file.type);
  if (VIDEO_CONTENT_TYPES.has(contentType)) {
    return contentType;
  }
  return VIDEO_CONTENT_TYPE_BY_EXTENSION[getExtension(file.name)] || "";
}

export function isOpenClawChatVideoFile(file: OpenClawChatVideoFileLike) {
  return Boolean(getOpenClawChatVideoContentType(file));
}

export function isOpenClawChatVideoAttachment(
  url: string,
  contentType?: string | null
) {
  return (
    normalizeContentType(contentType).startsWith("video/") ||
    VIDEO_EXTENSION_PATTERN.test(String(url || ""))
  );
}

export function normalizeOpenClawChatMediaAttachments(
  value: unknown
): OpenClawChatMediaAttachment[] {
  if (!Array.isArray(value)) {
    return [];
  }

  const result: OpenClawChatMediaAttachment[] = [];
  const seenUrls = new Set<string>();
  let videoCount = 0;

  for (const item of value) {
    if (!item || typeof item !== "object") {
      continue;
    }

    const raw = item as OpenClawChatAttachmentLike;
    const url = String(raw.url || "").trim();
    if (!url || !isHttpUrl(url) || seenUrls.has(url)) {
      continue;
    }

    const requestedType = String(raw.type || "").trim().toLowerCase();
    const rawContentType = normalizeContentType(
      raw.content_type || raw.contentType
    );
    const isVideo =
      requestedType === "video" ||
      isOpenClawChatVideoAttachment(url, rawContentType);

    if (isVideo && videoCount >= OPENCLAW_CHAT_VIDEO_MAX_FILES) {
      continue;
    }
    if (result.length >= OPENCLAW_CHAT_MEDIA_MAX_ATTACHMENTS) {
      break;
    }

    const type = isVideo ? "video" : "image";
    const fallbackContentType = isVideo
      ? VIDEO_CONTENT_TYPE_BY_EXTENSION[getExtension(url)] || "video/mp4"
      : "image/jpeg";
    const contentType =
      rawContentType.startsWith(`${type}/`) ? rawContentType : fallbackContentType;
    const name = String(raw.name || "")
      .trim()
      .slice(0, 200);

    result.push({
      type,
      url,
      ...(name ? { name } : {}),
      content_type: contentType,
    });
    seenUrls.add(url);
    if (isVideo) {
      videoCount += 1;
    }
  }

  return result;
}

export function buildOpenClawVideoDisplayText(
  attachments: OpenClawChatMediaAttachment[]
) {
  const videos = attachments.filter((attachment) => attachment.type === "video");
  if (videos.length === 0) {
    return "";
  }

  const lines = videos.map(
    (attachment, index) => `- ${attachment.name || `视频${index + 1}`}`
  );
  return `已上传视频：\n${lines.join("\n")}`;
}

export function buildOpenClawVideoReferenceText(
  attachments: OpenClawChatMediaAttachment[]
) {
  const videos = attachments.filter((attachment) => attachment.type === "video");
  if (videos.length === 0) {
    return "";
  }

  const lines = videos.map(
    (attachment, index) =>
      `- ${attachment.name || `视频${index + 1}`}: ${attachment.url}`
  );
  return [
    "用户上传了以下视频，请调用视频理解能力并按视频文件地址处理：",
    lines.join("\n"),
  ].join("\n");
}