fix(doc): strip duplicate title H1 on the JSONML doc-create path

internal/helpers/doc.go

@@ -335,6 +335,10 @@ func newDocCreateCommand(runner executor.Runner) *cobra.Command {
 					if err != nil {
 						return err
 					}
+					if stripped, ok := stripLeadingDuplicateTitleJSONML(jsonml, name); ok {
+						fmt.Fprintln(cmd.ErrOrStderr(), `note: 正文首个与 --name 相同的一级标题已自动移除（文档标题会单独渲染为页面标题，保留会出现两个标题）。`)
+						jsonml = stripped
+					}
 					createResult, err := docInvocationResult(cmd, runner, "doc", "create_document", params)
 					if err != nil {
 						return err
@@ -425,6 +429,83 @@ func stripLeadingDuplicateTitleHeading(content, name string) (string, bool) {
 	return strings.TrimLeft(rest, "\r\n"), true
 }
 
+// stripLeadingDuplicateTitleJSONML is the JSONML-path counterpart of
+// stripLeadingDuplicateTitleHeading. The rich-content path (--content-format
+// jsonml) writes the body verbatim via update_document, so a leading h1 node
+// whose text equals the document name renders the title twice — exactly the
+// problem the markdown path already guards against.
+//
+// jsonmlBody is the marshaled body produced by prepareDocJSONMLBody: a JSON
+// array, either a bare node list or wrapped as ["root", {}, ...nodes]. The
+// first content node is removed only when it is an h1 whose concatenated text
+// (trimmed, case-insensitive) exactly equals name. Any parse failure or
+// non-match returns the input untouched, so this never blocks a valid write.
+func stripLeadingDuplicateTitleJSONML(jsonmlBody, name string) (string, bool) {
+	name = strings.TrimSpace(name)
+	if name == "" {
+		return jsonmlBody, false
+	}
+	var body []any
+	if err := json.Unmarshal([]byte(jsonmlBody), &body); err != nil {
+		return jsonmlBody, false
+	}
+	// Locate the index of the first content node, skipping a ["root", {}, ...]
+	// wrapper if present.
+	first := 0
+	if len(body) >= 2 {
+		if tag, ok := body[0].(string); ok && tag == "root" {
+			first = 2
+		}
+	}
+	if first >= len(body) {
+		return jsonmlBody, false
+	}
+	node, ok := body[first].([]any)
+	if !ok || len(node) == 0 {
+		return jsonmlBody, false
+	}
+	tag, ok := node[0].(string)
+	if !ok || tag != "h1" {
+		return jsonmlBody, false
+	}
+	if !strings.EqualFold(strings.TrimSpace(jsonmlNodeText(node)), name) {
+		return jsonmlBody, false
+	}
+	body = append(body[:first], body[first+1:]...)
+	out, err := json.Marshal(body)
+	if err != nil {
+		return jsonmlBody, false
+	}
+	return string(out), true
+}
+
+// jsonmlNodeText concatenates all string leaves of a JSONML node, skipping the
+// leading tag name and any attributes object, so ["h1", {..}, ["span", {..},
+// ["span", {..}, "Title"]]] yields "Title".
+func jsonmlNodeText(node any) string {
+	switch v := node.(type) {
+	case string:
+		return v
+	case []any:
+		var b strings.Builder
+		for i, child := range v {
+			// element[0] is the tag name; a map at element[1] is attributes.
+			if i == 0 {
+				if _, isStr := child.(string); isStr {
+					continue
+				}
+			}
+			if _, isMap := child.(map[string]any); isMap {
+				continue
+			}
+			b.WriteString(jsonmlNodeText(child))
+		}
+		return b.String()
+	default:
+		return ""
+	}
+}
+
 func newDocUpdateCommand(runner executor.Runner) *cobra.Command {
 	cmd := &cobra.Command{
 		Use:               "update",

internal/helpers/doc_title_dedup_test.go

@@ -159,3 +159,110 @@ func TestDocCreateTitleOnlyContentOmitsMarkdown(t *testing.T) {
 		t.Errorf("markdown param = %v, want omitted", runner.last.Params["markdown"])
 	}
 }
+
+// TestDocCreateStripsDuplicateTitleJSONML verifies the end-to-end JSONML path:
+// `doc create --content-format jsonml` must drop a leading h1 whose text equals
+// --name before forwarding the body to update_document, otherwise the rich
+// document shows the page title twice.
+func TestDocCreateStripsDuplicateTitleJSONML(t *testing.T) {
+	runner := &docCommandRunner{responses: []map[string]any{{"nodeId": "NODE_X"}}}
+	root := newDocTestRoot(runner)
+
+	body := `{"jsonml":[["h1",{},"命令树参考"],["p",{},"正文"]]}`
+	_, errOut, err := executeDocCommand(t, root,
+		"create", "--name", "命令树参考",
+		"--content-format", "jsonml", "--content", body)
+	if err != nil {
+		t.Fatalf("execute: %v\nstderr:\n%s", err, errOut)
+	}
+	if len(runner.all) != 2 {
+		t.Fatalf("calls = %d, want 2 (create + update)", len(runner.all))
+	}
+	if runner.all[1].Tool != "update_document" {
+		t.Fatalf("second tool = %q, want update_document", runner.all[1].Tool)
+	}
+	got, _ := runner.all[1].Params["jsonml"].(string)
+	if strings.Contains(got, "命令树参考") {
+		t.Errorf("jsonml = %q, want duplicate h1 stripped", got)
+	}
+	if !strings.Contains(got, "正文") {
+		t.Errorf("jsonml = %q, want body content kept", got)
+	}
+	if !strings.Contains(errOut, "已自动移除") {
+		t.Errorf("stderr = %q, want a note about the removed heading", errOut)
+	}
+}
+
+// TestStripLeadingDuplicateTitleJSONML covers the JSONML-path counterpart: a
+// leading h1 node whose text equals the document name is removed; everything
+// else is left untouched.
+func TestStripLeadingDuplicateTitleJSONML(t *testing.T) {
+	tests := []struct {
+		name     string
+		body     string
+		docName  string
+		want     string
+		stripped bool
+	}{
+		{
+			name:     "root-wrapped duplicate h1 stripped",
+			body:     `["root",{},["h1",{},"晚会简报"],["p",{},"正文"]]`,
+			docName:  "晚会简报",
+			want:     `["root",{},["p",{},"正文"]]`,
+			stripped: true,
+		},
+		{
+			name:     "nested leaf text matches and stripped",
+			body:     `["root",{},["h1",{"uuid":"x"},["span",{"data-type":"text"},["span",{"data-type":"leaf"},"晚会简报"]]],["p",{},"正文"]]`,
+			docName:  "晚会简报",
+			want:     `["root",{},["p",{},"正文"]]`,
+			stripped: true,
+		},
+		{
+			name:     "bare body without root wrapper",
+			body:     `[["h1",{},"标题"],["p",{},"正文"]]`,
+			docName:  "标题",
+			want:     `[["p",{},"正文"]]`,
+			stripped: true,
+		},
+		{
+			name:     "case insensitive match",
+			body:     `["root",{},["h1",{},"Weekly REPORT"],["p",{},"x"]]`,
+			docName:  "weekly report",
+			want:     `["root",{},["p",{},"x"]]`,
+			stripped: true,
+		},
+		{
+			name:     "distinct heading kept",
+			body:     `["root",{},["h1",{},"背景"],["p",{},"正文"]]`,
+			docName:  "晚会简报",
+			want:     `["root",{},["h1",{},"背景"],["p",{},"正文"]]`,
+			stripped: false,
+		},
+		{
+			name:     "non-h1 leading node kept",
+			body:     `["root",{},["h2",{},"晚会简报"],["p",{},"正文"]]`,
+			docName:  "晚会简报",
+			want:     `["root",{},["h2",{},"晚会简报"],["p",{},"正文"]]`,
+			stripped: false,
+		},
+		{
+			name:     "invalid json untouched",
+			body:     `not json`,
+			docName:  "x",
+			want:     `not json`,
+			stripped: false,
+		},
+	}
+	for _, tc := range tests {
+		t.Run(tc.name, func(t *testing.T) {
+			got, ok := stripLeadingDuplicateTitleJSONML(tc.body, tc.docName)
+			if ok != tc.stripped {
+				t.Fatalf("stripped = %v, want %v", ok, tc.stripped)
+			}
+			if got != tc.want {
+				t.Fatalf("body = %q, want %q", got, tc.want)
+			}
+		})
+	}
+}