dsh-deepseek-import 0.1.1 → 0.1.2

package/README.md

@@ -54,7 +54,7 @@ Double-click **`dsh-import.cmd`** in the workspace (or drop JSON files onto it).
 
 - Lists your DSH workspaces; pick one by number
 - Each conversation becomes one session: title, order and body (including code blocks and tables) are preserved
-- **Attachments** (images, txt, pdf, docx…) that can be downloaded are saved under `<workspace>/.dsh-imported/<sessionId>/` with their original names; anything that cannot be downloaded degrades to `[附件未下载：name (size) 原链接]`
+- **Attachments** (images, txt, pdf, docx…): saved under `<workspace>/.dsh-imported/<sessionId>/` with their original names **only when the data source carries them**; anything that cannot be downloaded degrades to `[附件未下载：name (size) 原链接]`. DeepSeek's IndexedDB cache copy usually carries no attachments
 - **Thinking** goes to `thinking.md` in the same folder (never injected into the context)
 - **Edited / regenerated versions**: only the selected branch stays in the body, all versions go to `revisions.md`
 - **Interrupted outputs** keep only a numbered placeholder (`[中断输出 #27：未产生完整正文…]`); add `--keep-interrupted` to keep the partial text
@@ -188,6 +188,8 @@ Open the conversation and click **Capture chat**; then check the "来源" (sourc
 
 ## Known limitations
 
+- **Citations and attachments depend on the data source**: DeepSeek's IndexedDB cache copy keeps the body and thinking only; the `references` table and `files` field are usually absent. `references.md` and the attachment folder are therefore created **only when the source carries them** — their absence is a normal result, not a failure.
+
 - **Thinking is archived, not inlined**: DSH reasoning blocks carry provider signatures; faking them would poison later requests. Read `thinking.md`, or force inline with `--thinking`.
 - **Attachments are not added to DSH's attachment store**: they are written to disk and referenced by path. When no download URL is available only the name and size survive.
 - **Web-search citations**: the cache record has no `references` table, so `[reference:N]` markers stay as plain text with no links.

package/README.zh.md

@@ -57,7 +57,7 @@ DeepSeek 的长对话是**分页加载**的，每次只能看到一屏。扩展
 
 - 会列出 DSH 里的工作区，选序号决定导入到哪个工作区
 - 每个对话重建为一个会话，标题、顺序、正文（含代码块/表格等 Markdown）原样保留
-- **附件**（图片、txt、pdf、docx…）能下载的落盘到 `<工作区>/.dsh-imported/<sessionId>/`，保留原文件名；下载不到时退化成 `[附件未下载：文件名（大小） 原链接]`，由你本机匹配
+- **附件**（图片、txt、pdf、docx…）：**仅当数据源带附件时**才落盘到 `<工作区>/.dsh-imported/<sessionId>/`，保留原文件名；下载不到时退化成 `[附件未下载：文件名（大小） 原链接]`，由你本机匹配。DeepSeek 的 IndexedDB 缓存副本通常不含附件
 - **思考过程**写进同目录的 `thinking.md`（旁挂存档，不进入对话上下文）
 - **编辑/重新生成留下的多个版本**：正文只保留当前分支，全部版本写进同目录的 `revisions.md`
 - 改动 `workspace.json` 前自动备份为 `workspace.json.bak-<时间>`
@@ -225,6 +225,8 @@ node check-sessions.mjs
 
 ## 已知限制
 
+- **引用与附件取决于数据源**：DeepSeek 的 IndexedDB 缓存副本只保留正文和思考，`references` 表与 `files` 字段通常不在里面。因此 `references.md` 和附件目录**只在数据源带这两类数据时才会生成**，没有生成是正常结果，不是失败。想找回图片只能回原网页另存。
+
 - **思考过程旁挂**：写进 `thinking.md`，不进上下文。DSH 的 reasoning 块带 provider 签名，硬塞正文会污染后续请求；想看时直接翻文件，或加 `--thinking` 强制内联。
 - **附件不进 DSH 附件库**：DSH 的图片是内容寻址附件（`~/.dsh/attachments/v1/objects/...`），伪造引用有损坏历史的风险；这里改为落盘 + 路径引用，模型可以用工具读。拿不到下载地址时只留文件名与大小。
 - **正文完整性**：历史接口 `/api/v0/chat/history_messages` 返回的正文可能偏短；回答生成时的 `/api/v0/chat/completion` 是 `text/event-stream` 补丁流（`THINK` 片段=思考，其余片段=正文），扩展会边读边拼，生成完即写入捕获。所以**新回答**是完整的；很久以前的老回答以历史接口返回的内容为准。

package/dsh-import.mjs

@@ -424,7 +424,7 @@ function sidecarDir(workspacePath, sessionId) {
     const ignore = path.join(root, '.gitignore');
     if (!fs.existsSync(ignore)) {
       try {
-        fs.writeFileSync(ignore, '# DSH 导入存档（思考 / 引用 / 全文 / 附件）\n*\n');
+        fs.writeFileSync(ignore, '# DSH 导入存档（思考 / 全文 / 多版本；数据源带引用或附件时另存）\n*\n');
       } catch {
         /* 没有写权限就算了 */
       }

package/edge-extension/content.js

@@ -437,11 +437,31 @@
     try {
       const cached = await readCachedHistory();
       if (cached.ok && cached.messages?.length) {
-        pushCapture(cached.messages);
+        // 缓存里没有附件（DeepSeek 不把 files 写进 IndexedDB）。
+        // 当前屏幕上已渲染的消息能拿到 <img>，按「角色+正文完全相同」贴回去，
+        // 只做并集、不新增消息，避免重复。
+        let withFiles = cached.messages;
+        try {
+          const domMessages = domExtract().messages || [];
+          const byKey = new Map(withFiles.map((m) => [`${m.role}|${String(m.content || '').trim()}`, m]));
+          let attached = 0;
+          for (const domMessage of domMessages) {
+            if (!(domMessage.files || []).length) continue;
+            const target = byKey.get(`${domMessage.role}|${String(domMessage.content || '').trim()}`);
+            if (!target) continue;
+            target.files = [...(target.files || []), ...domMessage.files];
+            attached += domMessage.files.length;
+          }
+          if (attached) withFiles = [...withFiles];
+        } catch {
+          /* DOM 读不到就维持原样 */
+        }
+
+        pushCapture(withFiles);
         return {
           ok: true,
           source: 'cache',
-          messages: cached.messages,
+          messages: withFiles,
           rounds: 0,
           captures: 1,
           delta: 0,

package/edge-extension/merge.js

@@ -16,8 +16,25 @@
     return (message?.content || '').length + (message?.thinking || '').length;
   }
 
+  /** 同一消息的两份副本：正文取更全的，附件取并集 */
+  function unionFiles(a, b) {
+    const merged = [];
+    const seen = new Set();
+    for (const file of [...(a || []), ...(b || [])]) {
+      const key = file?.dataUrl ? String(file.dataUrl).slice(0, 64) : file?.url || file?.name;
+      if (!key || seen.has(key)) continue;
+      seen.add(key);
+      merged.push(file);
+    }
+    return merged;
+  }
+
   function richer(a, b) {
-    return score(b) > score(a) ? b : a;
+    if (!a) return b;
+    if (!b) return a;
+    const winner = score(b) > score(a) ? b : a;
+    const files = unionFiles(a.files, b.files);
+    return files.length ? { ...winner, files } : winner;
   }
 
   function keyOf(message) {

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "dsh-deepseek-import",
-  "version": "0.1.1",
+  "version": "0.1.2",
   "description": "DeepSeek web chat → DeepSeek Harness sessions (Edge extension + importer CLI)",
   "type": "module",
   "license": "MIT",