From 1e3bfa09d4e31e02d109ef2f4b41895dc89a78fe Mon Sep 17 00:00:00 2001
From: lihongyuan99 <64824864+lihongyuan99@users.noreply.github.com>
Date: Sun, 20 Sep 2026 20:26:48 +0800
Subject: [PATCH] fix(frontend): read web_fetch titles that start with blank
 lines or indented headings (#5560)

* fix(frontend): read web_fetch titles that start with blank lines or indented headings

* docs(frontend): describe the indented-code guard as it actually behaves

* fix(frontend): reject mixed code indentation in web-fetch titles

---------

Co-authored-by: Willem Jiang <willem.jiang@gmail.com>
---
 README.md                                     |  3 +
 frontend/AGENTS.md                            |  5 ++
 frontend/src/core/utils/markdown.ts           | 17 +++---
 .../tests/unit/core/utils/markdown.test.ts    | 60 +++++++++++++++++++
 4 files changed, 78 insertions(+), 7 deletions(-)
 create mode 100644 frontend/tests/unit/core/utils/markdown.test.ts

diff --git a/README.md b/README.md
index 8b2bd309f..4733d174e 100644
--- a/README.md
+++ b/README.md
@@ -1128,6 +1128,9 @@ empty list is forwarded and imposes no restriction of that kind. See the
 
 When using Tavily for `web_fetch`, extracted pages without a title use their URL
 as the heading; their content remains available to the agent.
+Chat tool-step titles accept leading blank lines and up to three spaces before
+a page's first H1 heading. Indented code, including mixed spaces and tabs, is
+not used as a title; the tool step falls back to the URL.
 Tavily search and fetch each read `api_key` from their own tool entry in
 `config.yaml`, falling back to `TAVILY_API_KEY` when omitted. Fetch does not
 reuse the search entry's key, so search can use a different provider. If you
diff --git a/frontend/AGENTS.md b/frontend/AGENTS.md
index 2d6f92529..8cd922271 100644
--- a/frontend/AGENTS.md
+++ b/frontend/AGENTS.md
@@ -83,6 +83,11 @@ More specific `AGENTS.md` files under `src/` contain the frontend sections split
 
 ## Code Style
 
+`core/utils/markdown.ts` reads web-fetch titles from the first nonblank line.
+Match zero to three literal spaces before `# ` without trimming indentation;
+mixed space/tab code blocks must fall back to the URL. Keep this local to title
+extraction rather than changing the shared streamdown fence parser.
+
 Custom Agent `display_name` is an optional Unicode UI label, edited in
 `AgentSettingsDialog`. Use it with a fallback to `name` for gallery/chat text;
 keep `name` for React identity, URLs, requests, and runtime `agent_name`.
diff --git a/frontend/src/core/utils/markdown.ts b/frontend/src/core/utils/markdown.ts
index fb22ebce9..fbef29001 100644
--- a/frontend/src/core/utils/markdown.ts
+++ b/frontend/src/core/utils/markdown.ts
@@ -1,10 +1,13 @@
+// Converter output can start with blank lines, and CommonMark allows up to three
+// literal spaces before an ATX heading. Do not trim code indentation into a title.
 export function extractTitleFromMarkdown(markdown: string) {
-  if (markdown.startsWith("# ")) {
-    let title = markdown.split("\n")[0]!.trim();
-    if (title.startsWith("# ")) {
-      title = title.slice(2).trim();
-    }
-    return title;
+  const firstLine = markdown.split("\n").find((line) => line.trim() !== "");
+  if (firstLine === undefined) {
+    return undefined;
   }
-  return undefined;
+  const headingPrefix = /^ {0,3}# /.exec(firstLine);
+  if (!headingPrefix) {
+    return undefined;
+  }
+  return firstLine.slice(headingPrefix[0].length).trim() || undefined;
 }
diff --git a/frontend/tests/unit/core/utils/markdown.test.ts b/frontend/tests/unit/core/utils/markdown.test.ts
new file mode 100644
index 000000000..140d64756
--- /dev/null
+++ b/frontend/tests/unit/core/utils/markdown.test.ts
@@ -0,0 +1,60 @@
+import { expect, test } from "@rstest/core";
+
+import { extractTitleFromMarkdown } from "@/core/utils/markdown";
+
+test("reads the title from a leading ATX heading", () => {
+  expect(extractTitleFromMarkdown("# Real Title\n\nbody")).toBe("Real Title");
+});
+
+test("skips blank lines before the first heading", () => {
+  expect(extractTitleFromMarkdown("\n# Real Title\n\nbody")).toBe("Real Title");
+  expect(extractTitleFromMarkdown("  \n\n# Real Title")).toBe("Real Title");
+});
+
+test("accepts the up-to-three-space indentation CommonMark allows", () => {
+  expect(extractTitleFromMarkdown("   # Real Title")).toBe("Real Title");
+});
+
+test("ignores an indented code block that starts with a hash", () => {
+  expect(extractTitleFromMarkdown("    # Not A Title")).toBeUndefined();
+  expect(extractTitleFromMarkdown("\t# Not A Title")).toBeUndefined();
+});
+
+test.each([" \t", "  \t", "   \t"])(
+  "does not treat mixed indentation %j as a heading",
+  (indent) => {
+    expect(
+      extractTitleFromMarkdown(`${indent}# Code comment\n# Later heading`),
+    ).toBeUndefined();
+    expect(
+      extractTitleFromMarkdown(`\n  \n${indent}# Code comment`),
+    ).toBeUndefined();
+  },
+);
+
+test.each(["", " ", "  ", "   "])(
+  "accepts a heading with %j indentation and CRLF line endings",
+  (indent) => {
+    expect(extractTitleFromMarkdown(`\r\n${indent}# Real Title\r\nbody`)).toBe(
+      "Real Title",
+    );
+  },
+);
+
+test("ignores headings that are not level 1", () => {
+  expect(extractTitleFromMarkdown("## Section")).toBeUndefined();
+  expect(extractTitleFromMarkdown("#NoSpace")).toBeUndefined();
+});
+
+test("does not report an empty heading as a title", () => {
+  expect(extractTitleFromMarkdown("# \n\nbody")).toBeUndefined();
+  expect(extractTitleFromMarkdown("#")).toBeUndefined();
+});
+
+test("returns undefined when the document has no content", () => {
+  expect(extractTitleFromMarkdown("")).toBeUndefined();
+  expect(extractTitleFromMarkdown("   \n  ")).toBeUndefined();
+  expect(
+    extractTitleFromMarkdown("Plain text with no heading"),
+  ).toBeUndefined();
+});