feat: chat uploads, vision turn isolation, and MindSpace agent improvements

Add chat file/image upload UX, attachment proxying, vision thumbnails, and per-turn image scoping so agents only use the current upload. Extend MindSpace asset context, billing token state, OA/scenario verify scripts, and related runtime config.

Co-authored-by: Cursor <cursoragent@cursor.com>

src/utils/message.ts

@@ -1,4 +1,4 @@
-import type { Message, MessageContent } from '../types';
+import type { ChatFileAttachment, Message, MessageContent } from '../types';
 import { mergeMessageContent } from '../../message-stream.mjs';
 import { mergeConversationSnapshot as mergeConversationSnapshotCore } from '../../chat-finish-sync.mjs';
 import { deriveUserFacingText, deriveAssistantFacingText } from '../../conversation-display.mjs';
@@ -6,8 +6,11 @@ import { stripUserAddressPrefix } from './userAddress';
 
 const IMAGE_URL_LINE_RE = /^\[图片\d+]: (.+)$/;
 const IMAGE_URL_LINES_RE = /\n*\[图片\d+]: [^\n]+/g;
+const FILE_ATTACHMENT_LINE_RE = /^\[文件\d+: ([^\]]+)\]: (.+)$/;
+const FILE_ATTACHMENT_LINES_RE = /\n*\[文件\d+: [^\]]+\]: [^\n]+/g;
 const USER_IDENTITY_BLOCK_RE = /^\[用户身份\][\s\S]*?(?:\n{2,}|$)/;
 const TKMIND_VISION_NOTE_RE = /\n*【TKMind 图片分析结果[\s\S]*$/;
+const TKMIND_ATTACHMENT_NOTE_RE = /\n*【TKMind 附件分析结果[\s\S]*$/;
 
 function normalizeImageUrl(url: string): string {
   const value = String(url ?? '').trim();
@@ -41,10 +44,19 @@ function stripImageUrlLines(text: string): string {
   return String(text ?? '')
     .replace(USER_IDENTITY_BLOCK_RE, '')
     .replace(TKMIND_VISION_NOTE_RE, '')
+    .replace(TKMIND_ATTACHMENT_NOTE_RE, '')
     .replace(IMAGE_URL_LINES_RE, '')
+    .replace(FILE_ATTACHMENT_LINES_RE, '')
     .trim();
 }
 
+function formatFileAttachmentsForAgentText(attachments: ChatFileAttachment[]): string {
+  return attachments
+    .filter((item) => item?.downloadUrl?.trim())
+    .map((item, index) => `[文件${index + 1}: ${item.filename}]: ${item.downloadUrl.trim()}`)
+    .join('\n');
+}
+
 function parseImageUrlsFromText(text: string): string[] {
   const urls: string[] = [];
   for (const line of text.split('\n')) {
@@ -82,6 +94,7 @@ export function normalizeUserMessageForApi(message: Message): Message {
   if (message.role !== 'user') return message;
 
   const imageUrls = getImageUrls(message);
+  const fileAttachments = getFileAttachments(message);
   const displayText =
     message.metadata.displayText ??
     stripImageUrlLines(
@@ -97,12 +110,17 @@ export function normalizeUserMessageForApi(message: Message): Message {
     .join('\n')
     .trim();
 
-  if (imageUrls.length === 0 && message.content.every((item) => item.type === 'text')) {
+  if (
+    imageUrls.length === 0 &&
+    fileAttachments.length === 0 &&
+    message.content.every((item) => item.type === 'text')
+  ) {
     return message;
   }
 
   const imageText = formatImageUrlsForAgentText(imageUrls);
-  const agentText = [stripImageUrlLines(textOnly), imageText]
+  const fileText = formatFileAttachmentsForAgentText(fileAttachments);
+  const agentText = [stripImageUrlLines(textOnly), imageText, fileText]
     .filter(Boolean)
     .join('\n\n');
 
@@ -113,6 +131,7 @@ export function normalizeUserMessageForApi(message: Message): Message {
       ...message.metadata,
       displayText,
       ...(imageUrls.length ? { imageUrls } : {}),
+      ...(fileAttachments.length ? { fileAttachments } : {}),
     },
   };
 }
@@ -125,6 +144,7 @@ export function buildUserMessage(
     displayText?: string;
     imageUrls?: string[];
     previewImageUrls?: string[];
+    fileAttachments?: ChatFileAttachment[];
   },
 ): Message {
   const displayText = stripImageUrlLines(options?.displayText ?? text);
@@ -132,9 +152,13 @@ export function buildUserMessage(
   const previewImageUrls = (options?.previewImageUrls ?? []).filter(
     (value) => typeof value === 'string' && value.trim(),
   );
+  const fileAttachments = (options?.fileAttachments ?? []).filter(
+    (item) => item?.downloadUrl?.trim() && item?.filename?.trim(),
+  );
   const baseAgentText = options?.agentText ?? text;
   const imageText = formatImageUrlsForAgentText(imageUrls);
-  const agentText = [baseAgentText.trim(), imageText].filter(Boolean).join('\n\n');
+  const fileText = formatFileAttachmentsForAgentText(fileAttachments);
+  const agentText = [baseAgentText.trim(), imageText, fileText].filter(Boolean).join('\n\n');
 
   return {
     id: options?.id ?? crypto.randomUUID(),
@@ -147,6 +171,7 @@ export function buildUserMessage(
       displayText,
       ...(imageUrls.length ? { imageUrls } : {}),
       ...(previewImageUrls.length ? { previewImageUrls } : {}),
+      ...(fileAttachments.length ? { fileAttachments } : {}),
     },
   };
 }
@@ -199,6 +224,7 @@ export function getDisplayText(message: Message): string {
 export function shouldShowChatMessage(message: Message): boolean {
   if (getDisplayText(message).trim()) return true;
   if (message.role === 'user' && getImageUrls(message).length > 0) return true;
+  if (message.role === 'user' && getFileAttachments(message).length > 0) return true;
   if (getThinking(message)) return true;
   if (getSystemNotificationText(message)) return true;
   if (message.role !== 'assistant') return false;
@@ -230,7 +256,10 @@ export function getVisibleText(message: Message): string {
 }
 
 export function getImageUrls(message: Message): string[] {
-  const fromMetadata = message.metadata.imageUrls?.filter(
+  const fromMetadata = (
+    message.metadata.imageUrls
+    ?? message.metadata.archivedImageUrls
+  )?.filter(
     (url) => typeof url === 'string' && url.trim(),
   );
   if (fromMetadata?.length) return fromMetadata.map(normalizeImageUrl);
@@ -245,9 +274,36 @@ export function getImageUrls(message: Message): string[] {
   return parseImageUrlsFromText(text).map(normalizeImageUrl);
 }
 
+export function getFileAttachments(message: Message): ChatFileAttachment[] {
+  const fromMetadata = message.metadata.fileAttachments?.filter(
+    (item) => item?.downloadUrl?.trim() && item?.filename?.trim(),
+  );
+  if (fromMetadata?.length) return fromMetadata;
+
+  const text = message.content
+    .filter((item): item is Extract<MessageContent, { type: 'text' }> => item.type === 'text')
+    .map((item) => item.text)
+    .join('\n');
+  const parsed: ChatFileAttachment[] = [];
+  for (const line of text.split('\n')) {
+    const match = line.trim().match(FILE_ATTACHMENT_LINE_RE);
+    if (!match?.[1] || !match?.[2]) continue;
+    parsed.push({
+      assetId: '',
+      filename: match[1],
+      downloadUrl: match[2],
+      mimeType: 'application/octet-stream',
+    });
+  }
+  return parsed;
+}
+
 export function getRenderableImageUrls(message: Message): Array<{ src: string; href: string }> {
   const imageUrls = getImageUrls(message);
-  const previewImageUrls = message.metadata.previewImageUrls?.filter(
+  const previewImageUrls = (
+    message.metadata.previewImageUrls
+    ?? message.metadata.archivedPreviewImageUrls
+  )?.filter(
     (url) => typeof url === 'string' && url.trim(),
   ) ?? [];
   return imageUrls.map((href, index) => ({