feat(maestro): show image previews in user message bubbles

Uploaded images now render as thumbnails in the user bubble instead of
showing only the server file path as text.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

web/src/structured.ts

@@ -253,6 +253,8 @@ export function createStructuredConnection(
   let activityTimer: number | null = null;
   const promptQueue: QueuedPrompt[] = [];
   let promptRequestPending = false;
+  /** Server path → object URL for uploaded images, so user bubbles can show previews. */
+  const imageObjectUrls = new Map<string, string>();
 
   // Within 64px, wheel and touch imprecision should still count as following.
   const FOLLOW_BOTTOM_THRESHOLD_PX = 64;
@@ -383,6 +385,9 @@ export function createStructuredConnection(
     }
   }
 
+  /** Pattern matching `Attached file "name": "path"` references appended by dispatchPrompt. */
+  const attachedFileRe = /Attached file "(?:[^"\\]|\\.)*": "((?:[^"\\]|\\.)*)"/g;
+
   function buildRowElement(row: TranscriptRow): HTMLElement | null {
     const el = buildTranscriptRow(
       row,
@@ -399,6 +404,39 @@ export function createStructuredConnection(
     if (el.querySelector(".st-tool-head")?.getAttribute("aria-expanded") !== "true") {
       expandedRowIds.delete(row.id);
     }
+    // Inject image previews into user message bubbles and strip the path text.
+    if (row.kind === "user" && imageObjectUrls.size > 0) {
+      const data = row.data as { content?: { type: string; text?: string } } | undefined;
+      const text = data?.content?.type === "text" ? data.content.text ?? "" : "";
+      const body = el.querySelector(".st-body");
+      if (body && text) {
+        const images: string[] = [];
+        attachedFileRe.lastIndex = 0;
+        let match: RegExpExecArray | null;
+        while ((match = attachedFileRe.exec(text)) !== null) {
+          const serverPath = JSON.parse(`"${match[1]}"`);
+          const objectUrl = imageObjectUrls.get(serverPath);
+          if (objectUrl) images.push(objectUrl);
+        }
+        if (images.length > 0) {
+          // Strip attachment reference lines from the displayed text.
+          const userText = body.querySelector(".st-user-text");
+          if (userText) {
+            const cleaned = text.replace(/\n\nAttached file "[^"]*": "[^"]*"/g, "").trim();
+            userText.textContent = cleaned;
+            if (!cleaned) userText.remove();
+          }
+          for (const src of images) {
+            const img = document.createElement("img");
+            img.className = "st-user-image";
+            img.src = src;
+            img.alt = "Uploaded image";
+            img.loading = "lazy";
+            body.appendChild(img);
+          }
+        }
+      }
+    }
     return el;
   }
 
@@ -991,6 +1029,9 @@ export function createStructuredConnection(
       for (const attachment of submittedAttachments) {
         if (disposed) return;
         if (!attachment.path) attachment.path = (await uploadStructuredFile(sessionId, attachment.file)).path;
+        if (attachment.path && attachment.file.type.startsWith("image/")) {
+          imageObjectUrls.set(attachment.path, URL.createObjectURL(attachment.file));
+        }
       }
       if (disposed) return;
       const references = submittedAttachments.map(({ file, path }) =>
@@ -1230,6 +1271,8 @@ export function createStructuredConnection(
       closeCurrentSocket();
       expandedRowIds.clear();
       expandedGroupIds.clear();
+      for (const url of imageObjectUrls.values()) URL.revokeObjectURL(url);
+      imageObjectUrls.clear();
       scrollEl.removeEventListener("scroll", onTranscriptScroll);
       scrollEl.removeEventListener("click", onTranscriptClick, true);
       scrollEl.removeEventListener("wheel", stopSmoothFollow);

web/src/structuredRender.test.ts

@@ -315,8 +315,8 @@ describe("buildTranscriptRow: tool diffs", () => {
     (el.querySelector(".st-tool-head") as HTMLButtonElement).click();
     const diff = el.querySelector(".st-diff") as HTMLElement;
     expect(diff).not.toBeNull();
-    expect(diff.querySelector(".st-ln-add")!.textContent).toBe("+ c");
-    expect(diff.querySelector(".st-ln-del")!.textContent).toBe("- b");
+    expect(diff.querySelector(".st-ln-add")!.textContent).toBe("c");
+    expect(diff.querySelector(".st-ln-del")!.textContent).toBe("b");
   });
 
   it("gives the diff its own horizontally-scrolling container", () => {

web/src/structuredRender.ts

@@ -223,16 +223,15 @@ function buildDiffBlock(diff: { path: string; oldText?: string | null; newText:
   wrap.appendChild(label);
   const scroller = el("div", "st-diff");
   const pre = document.createElement("pre");
-  const lines = computeDiffLines(diff.oldText, diff.newText);
-  const prefixes: Record<string, string> = { add: "+ ", del: "- ", ctx: "  ", hunk: "", eof: "" };
-  for (const line of lines) {
+  for (const line of computeDiffLines(diff.oldText, diff.newText)) {
     const span = document.createElement("span");
     span.className = `st-ln st-ln-${line.kind}`;
-    span.textContent = (prefixes[line.kind] ?? "") + line.text;
+    span.textContent = line.text;
     pre.appendChild(span);
+    pre.appendChild(document.createTextNode("\n"));
   }
   scroller.appendChild(pre);
-  makeCopyable(scroller, lines.map((l) => (prefixes[l.kind] ?? "") + l.text).join("\n"), "diff");
+  makeCopyable(scroller, pre.textContent ?? "", "diff");
   wrap.appendChild(scroller);
   return wrap;
 }

web/src/styles.css

@@ -2380,6 +2380,15 @@ input.leaf-rename:focus {
   overflow-wrap: anywhere;
 }
 
+.st-user-image {
+  display: block;
+  max-width: min(100%, 480px);
+  max-height: 360px;
+  margin-top: var(--sp-2);
+  border-radius: var(--r-1);
+  object-fit: contain;
+}
+
 /* agent prose: no card, no background — plain text is the point */
 .st-row-agent .st-body p {
   margin: 0 0 var(--sp-2);
@@ -2965,7 +2974,6 @@ input.leaf-rename:focus {
   display: block;
   white-space: pre;
   padding: 0 var(--sp-3);
-  line-height: 1.4;
 }
 
 .st-ln-add {