feat: support whiteboard file inputs in docs XML

2026-08-03 08:32:46 +08:00 · 2026-07-07 19:44:13 +08:00
parent f98dbfe247
commit d8cb216f92
4 changed files with 407 additions and 2 deletions
--- a/shortcuts/doc/html5_block_resources.go
+++ b/shortcuts/doc/html5_block_resources.go
@@ -27,12 +27,17 @@ const (
 	html5BlockDataAttr        = "data"
 	html5BlockReferenceRoot   = "doc-fetch-resources"
 	html5BlockReferenceMaxRaw = 1024
+
+	whiteboardTag      = "whiteboard"
+	whiteboardTypeAttr = "type"
+	whiteboardPathAttr = "path"
 )

 var (
 	html5BlockStartTagPattern = regexp.MustCompile(`(?is)<html5-block\b[^>]*>`)
 	html5BlockElementPattern  = regexp.MustCompile(`(?is)<html5-block\b[^>]*>(.*?)</html5-block>`)
 	html5BlockSafeNamePattern = regexp.MustCompile(`^[A-Za-z0-9._-]+$`)
+	whiteboardElementPattern  = regexp.MustCompile(`(?is)<whiteboard\b[^>]*(?:/>|>.*?</whiteboard>)`)
 )

 type html5BlockReferenceEntry struct {
@@ -58,6 +63,11 @@ type html5BlockStartTag struct {
 	SelfClosing bool
 }

+type whiteboardStartTag struct {
+	Attrs       []html5BlockAttr
+	SelfClosing bool
+}
+
 func buildCreateBodyWithHTML5ReferenceMap(runtime *common.RuntimeContext) (map[string]interface{}, error) {
 	body := buildCreateBody(runtime)
 	if runtime.Str("content") == "" && !runtime.Changed("reference-map") {
@@ -115,7 +125,11 @@ func prepareDocsV2WriteInput(runtime *common.RuntimeContext, input docsV2WriteIn
 		return docsV2WriteInput{}, err
 	}

-	content, html5RefMap, err := prepareHTML5BlockWriteContent(runtime, runtime.Str("doc-format"), input.Content, html5RefMap)
+	content, err := prepareWhiteboardWriteContent(runtime, runtime.Str("doc-format"), input.Content)
+	if err != nil {
+		return docsV2WriteInput{}, err
+	}
+	content, html5RefMap, err = prepareHTML5BlockWriteContent(runtime, runtime.Str("doc-format"), content, html5RefMap)
 	if err != nil {
 		return docsV2WriteInput{}, err
 	}
@@ -232,6 +246,213 @@ func prepareHTML5BlockWriteContent(runtime *common.RuntimeContext, format string
 	return out, compactReferenceMap(refMap), nil
 }

+func prepareWhiteboardWriteContent(runtime *common.RuntimeContext, format string, content string) (string, error) {
+	if !strings.Contains(content, "<whiteboard") {
+		return content, nil
+	}
+
+	rewrite := func(segment string) (string, error) {
+		return rewriteWhiteboardFileRefs(runtime, segment)
+	}
+
+	if strings.TrimSpace(format) != "markdown" {
+		return rewrite(content)
+	}
+
+	var (
+		out strings.Builder
+		err error
+	)
+	out.WriteString(applyOutsideCodeFences(content, func(segment string) string {
+		if err != nil {
+			return segment
+		}
+		outSegment, rewriteErr := rewrite(segment)
+		if rewriteErr != nil {
+			err = rewriteErr
+			return segment
+		}
+		return outSegment
+	}))
+	if err != nil {
+		return "", err
+	}
+	return out.String(), nil
+}
+
+func rewriteWhiteboardFileRefs(runtime *common.RuntimeContext, content string) (string, error) {
+	var rewriteErrs []error
+	out := whiteboardElementPattern.ReplaceAllStringFunc(content, func(raw string) string {
+		rewritten, err := rewriteWhiteboardFileRef(runtime, raw)
+		if err != nil {
+			rewriteErrs = append(rewriteErrs, err)
+			return raw
+		}
+		return rewritten
+	})
+	if len(rewriteErrs) > 0 {
+		return "", aggregateWhiteboardRewriteErrors(rewriteErrs)
+	}
+	return out, nil
+}
+
+func rewriteWhiteboardFileRef(runtime *common.RuntimeContext, raw string) (string, error) {
+	startRaw, body, _, ok := splitWhiteboardElement(raw)
+	if !ok {
+		return raw, nil
+	}
+	tag, err := parseWhiteboardStartTag(startRaw)
+	if err != nil {
+		return "", common.ValidationErrorf("invalid whiteboard tag: %v", err).WithParam("whiteboard")
+	}
+
+	pathValue, hasPath := tag.attr(whiteboardPathAttr)
+	bodyPath, hasBodyPath := whiteboardBodyPathRef(body)
+	if !hasPath && !hasBodyPath {
+		return raw, nil
+	}
+	if hasPath && strings.TrimSpace(body) != "" {
+		return "", common.ValidationErrorf("whiteboard cannot contain both path and inline content").WithParam("whiteboard")
+	}
+	if hasPath && hasBodyPath {
+		return "", common.ValidationErrorf("whiteboard cannot contain both path and @file body").WithParam("whiteboard")
+	}
+
+	typRaw, ok := tag.attr(whiteboardTypeAttr)
+	if !ok || strings.TrimSpace(typRaw) == "" {
+		return "", common.ValidationErrorf("whiteboard file input requires type=\"svg\", type=\"mermaid\", or type=\"plantuml\"").WithParam("type")
+	}
+	typ, ok := canonicalWhiteboardFileType(typRaw)
+	if !ok {
+		return "", common.ValidationErrorf("whiteboard file input only supports type=\"svg\", type=\"mermaid\", or type=\"plantuml\", got %q", typRaw).WithParam("type")
+	}
+
+	if hasBodyPath {
+		pathValue = bodyPath
+	}
+	data, err := readWhiteboardPath(runtime, pathValue, typ)
+	if err != nil {
+		return "", err
+	}
+
+	tag.setAttr(whiteboardTypeAttr, typ)
+	tag.removeAttrs(whiteboardPathAttr)
+	return tag.render(false) + whiteboardContentForType(typ, data) + "</" + whiteboardTag + ">", nil
+}
+
+func splitWhiteboardElement(raw string) (startTag string, body string, selfClosing bool, ok bool) {
+	trimmed := strings.TrimSpace(raw)
+	selfClosing = strings.HasSuffix(trimmed, "/>")
+	if selfClosing {
+		return raw, "", true, true
+	}
+	startEnd := strings.Index(raw, ">")
+	if startEnd < 0 {
+		return "", "", false, false
+	}
+	endStart := strings.LastIndex(strings.ToLower(raw), "</whiteboard>")
+	if endStart < 0 || endStart < startEnd {
+		return "", "", false, false
+	}
+	return raw[:startEnd+1], raw[startEnd+1 : endStart], false, true
+}
+
+func whiteboardBodyPathRef(body string) (string, bool) {
+	trimmed := strings.TrimSpace(body)
+	if !strings.HasPrefix(trimmed, "@") || strings.HasPrefix(trimmed, "@@") {
+		return "", false
+	}
+	if strings.ContainsAny(trimmed, "\r\n") {
+		return "", false
+	}
+	return trimmed, true
+}
+
+func canonicalWhiteboardFileType(raw string) (string, bool) {
+	switch strings.ToLower(strings.TrimSpace(raw)) {
+	case "svg":
+		return "svg", true
+	case "mermaid":
+		return "mermaid", true
+	case "plantuml":
+		return "plantuml", true
+	default:
+		return "", false
+	}
+}
+
+func readWhiteboardPath(runtime *common.RuntimeContext, pathValue string, typ string) (string, error) {
+	pathRaw := strings.TrimSpace(pathValue)
+	if !strings.HasPrefix(pathRaw, "@") {
+		return "", common.ValidationErrorf("whiteboard %s path %q must start with @, for example @diagram.%s", typ, pathValue, exampleWhiteboardExt(typ)).WithParam("path")
+	}
+	relPath := strings.TrimSpace(strings.TrimPrefix(pathRaw, "@"))
+	if relPath == "" {
+		return "", common.ValidationErrorf("whiteboard %s path cannot be empty after @", typ).WithParam("path")
+	}
+	clean := filepath.Clean(relPath)
+	if filepath.IsAbs(clean) || clean == "." || clean == ".." || strings.HasPrefix(clean, ".."+string(filepath.Separator)) {
+		return "", common.ValidationErrorf("whiteboard %s path %q must be a relative path within the current working directory", typ, pathValue).WithParam("path")
+	}
+	if !whiteboardExtAllowed(typ, strings.ToLower(filepath.Ext(clean))) {
+		return "", common.ValidationErrorf("whiteboard %s path %q must point to a %s file", typ, pathValue, whiteboardExtList(typ)).WithParam("path")
+	}
+	data, err := cmdutil.ReadInputFile(runtime.FileIO(), clean)
+	if err != nil {
+		return "", common.ValidationErrorf("whiteboard %s path %q cannot be read from the current working directory; check that the file exists relative to where lark-cli is running: %v", typ, clean, err).WithParam("path").WithCause(err)
+	}
+	return string(data), nil
+}
+
+func whiteboardExtAllowed(typ string, ext string) bool {
+	for _, allowed := range whiteboardAllowedExts(typ) {
+		if ext == allowed {
+			return true
+		}
+	}
+	return false
+}
+
+func whiteboardAllowedExts(typ string) []string {
+	switch typ {
+	case "svg":
+		return []string{".svg"}
+	case "mermaid":
+		return []string{".mermaid", ".mmd"}
+	case "plantuml":
+		return []string{".plantuml", ".puml", ".pu", ".uml"}
+	default:
+		return nil
+	}
+}
+
+func whiteboardExtList(typ string) string {
+	return strings.Join(whiteboardAllowedExts(typ), ", ")
+}
+
+func exampleWhiteboardExt(typ string) string {
+	exts := whiteboardAllowedExts(typ)
+	if len(exts) == 0 {
+		return "txt"
+	}
+	return strings.TrimPrefix(exts[0], ".")
+}
+
+func whiteboardContentForType(typ string, data string) string {
+	if typ == "svg" {
+		return data
+	}
+	return escapeXMLText(data)
+}
+
+func aggregateWhiteboardRewriteErrors(rewriteErrs []error) error {
+	messages := make([]string, 0, len(rewriteErrs))
+	for _, err := range rewriteErrs {
+		messages = append(messages, err.Error())
+	}
+	return common.ValidationErrorf("whiteboard file input failed: %s", strings.Join(messages, "; ")).WithParam("whiteboard").WithCause(errors.Join(rewriteErrs...))
+}
+
 func validateHTML5BlockWriteElementBodies(format string, content string) error {
 	validateSegment := func(segment string) error {
 		matches := html5BlockElementPattern.FindAllStringSubmatchIndex(segment, -1)
@@ -621,6 +842,34 @@ func parseHTML5BlockStartTag(raw string) (html5BlockStartTag, error) {
 	return html5BlockStartTag{}, fmt.Errorf("missing start element") //nolint:forbidigo // intermediate parse helper; callers wrap with typed validation errors.
 }

+func parseWhiteboardStartTag(raw string) (whiteboardStartTag, error) {
+	trimmed := strings.TrimSpace(raw)
+	selfClosing := strings.HasSuffix(trimmed, "/>")
+	decoder := xml.NewDecoder(strings.NewReader(raw))
+	for {
+		tok, err := decoder.Token()
+		if err != nil {
+			if errors.Is(err, io.EOF) {
+				break
+			}
+			return whiteboardStartTag{}, err
+		}
+		start, ok := tok.(xml.StartElement)
+		if !ok {
+			continue
+		}
+		if start.Name.Local != whiteboardTag {
+			return whiteboardStartTag{}, fmt.Errorf("expected <%s>, got <%s>", whiteboardTag, start.Name.Local) //nolint:forbidigo // intermediate parse helper; callers wrap with typed validation errors.
+		}
+		attrs := make([]html5BlockAttr, 0, len(start.Attr))
+		for _, attr := range start.Attr {
+			attrs = append(attrs, html5BlockAttr{Name: attr.Name.Local, Value: attr.Value})
+		}
+		return whiteboardStartTag{Attrs: attrs, SelfClosing: selfClosing}, nil
+	}
+	return whiteboardStartTag{}, fmt.Errorf("missing start element") //nolint:forbidigo // intermediate parse helper; callers wrap with typed validation errors.
+}
+
 func (t html5BlockStartTag) attr(name string) (string, bool) {
 	for _, attr := range t.Attrs {
 		if attr.Name == name {
@@ -630,6 +879,15 @@ func (t html5BlockStartTag) attr(name string) (string, bool) {
 	return "", false
 }

+func (t whiteboardStartTag) attr(name string) (string, bool) {
+	for _, attr := range t.Attrs {
+		if attr.Name == name {
+			return attr.Value, true
+		}
+	}
+	return "", false
+}
+
 func (t html5BlockStartTag) hasAttr(name string) bool {
 	_, ok := t.attr(name)
 	return ok
@@ -650,6 +908,31 @@ func (t *html5BlockStartTag) removeAttrs(names ...string) {
 	t.Attrs = attrs
 }

+func (t *whiteboardStartTag) removeAttrs(names ...string) {
+	remove := make(map[string]struct{}, len(names))
+	for _, name := range names {
+		remove[name] = struct{}{}
+	}
+	attrs := t.Attrs[:0]
+	for _, attr := range t.Attrs {
+		if _, ok := remove[attr.Name]; ok {
+			continue
+		}
+		attrs = append(attrs, attr)
+	}
+	t.Attrs = attrs
+}
+
+func (t *whiteboardStartTag) setAttr(name string, value string) {
+	for i, attr := range t.Attrs {
+		if attr.Name == name {
+			t.Attrs[i].Value = value
+			return
+		}
+	}
+	t.Attrs = append(t.Attrs, html5BlockAttr{Name: name, Value: value})
+}
+
 func (t html5BlockStartTag) render(selfClosing bool) string {
 	var b strings.Builder
 	b.WriteByte('<')
@@ -674,6 +957,25 @@ func (t html5BlockStartTag) render(selfClosing bool) string {
 	return b.String()
 }

+func (t whiteboardStartTag) render(selfClosing bool) string {
+	var b strings.Builder
+	b.WriteByte('<')
+	b.WriteString(whiteboardTag)
+	for _, attr := range t.Attrs {
+		b.WriteByte(' ')
+		b.WriteString(attr.Name)
+		b.WriteString(`="`)
+		b.WriteString(escapeXMLAttr(attr.Value))
+		b.WriteByte('"')
+	}
+	if selfClosing {
+		b.WriteString("/>")
+	} else {
+		b.WriteByte('>')
+	}
+	return b.String()
+}
+
 func escapeXMLAttr(value string) string {
 	var b strings.Builder
 	for _, r := range value {
@@ -694,3 +996,18 @@ func escapeXMLAttr(value string) string {
 	}
 	return b.String()
 }
+
+func escapeXMLText(value string) string {
+	var b strings.Builder
+	for _, r := range value {
+		switch r {
+		case '&':
+			b.WriteString("&amp;")
+		case '<':
+			b.WriteString("&lt;")
+		default:
+			b.WriteRune(r)
+		}
+	}
+	return b.String()
+}
--- a/shortcuts/doc/html5_block_resources_test.go
+++ b/shortcuts/doc/html5_block_resources_test.go
@@ -116,6 +116,61 @@ func TestDocsCreateV2HTML5BlockReferenceMapFromPath(t *testing.T) {
 	}
 }

+func TestDocsCreateV2WhiteboardFileInputs(t *testing.T) {
+	dir := t.TempDir()
+	cmdutil.TestChdir(t, dir)
+	files := map[string]string{
+		"diagram.svg":   `<svg viewBox="0 0 10 10"><text>A</text></svg>`,
+		"flow.mmd":      "flowchart TD\nA --> B",
+		"sequence.puml": "@startuml\nAlice -> Bob: hi\n@enduml",
+	}
+	for name, content := range files {
+		if err := os.WriteFile(name, []byte(content), 0o600); err != nil {
+			t.Fatalf("WriteFile(%s) error: %v", name, err)
+		}
+	}
+
+	f, stdout, _, reg := cmdutil.TestFactory(t, docsCreateTestConfig(t, ""))
+	stub := registerDocsAIStub(reg, "POST", "/open-apis/docs_ai/v1/documents", map[string]interface{}{
+		"document": map[string]interface{}{
+			"document_id": "doxcn_new_doc",
+			"revision_id": float64(1),
+		},
+	})
+
+	err := runDocsCreateShortcut(t, f, stdout, []string{
+		"+create",
+		"--api-version", "v2",
+		"--content", strings.Join([]string{
+			`<whiteboard type="svg" path="@diagram.svg"></whiteboard>`,
+			`<whiteboard type="mermaid">@flow.mmd</whiteboard>`,
+			`<whiteboard type="plantUML" path="@sequence.puml"/>`,
+		}, "\n"),
+		"--as", "user",
+	})
+	if err != nil {
+		t.Fatalf("unexpected error: %v", err)
+	}
+
+	body := decodeRequestBody(t, stub.CapturedBody)
+	got := body["content"].(string)
+	for _, want := range []string{
+		`<whiteboard type="svg"><svg viewBox="0 0 10 10"><text>A</text></svg></whiteboard>`,
+		"<whiteboard type=\"mermaid\">flowchart TD\nA --> B</whiteboard>",
+		"<whiteboard type=\"plantuml\">@startuml\nAlice -> Bob: hi\n@enduml</whiteboard>",
+	} {
+		if !strings.Contains(got, want) {
+			t.Fatalf("content missing %q:\n%s", want, got)
+		}
+	}
+	if strings.Contains(got, `path="@`) {
+		t.Fatalf("content still contains whiteboard path attr: %s", got)
+	}
+	if _, ok := body["reference_map"]; ok {
+		t.Fatalf("whiteboard file input must not create reference_map: %#v", body)
+	}
+}
+
 func findDocsTestFlag(flags []common.Flag, name string) common.Flag {
 	for _, flag := range flags {
 		if flag.Name == name {
@@ -407,6 +462,35 @@ func TestDocsCreateV2HTML5BlockPathReadFailure(t *testing.T) {
 	}
 }

+func TestDocsCreateV2WhiteboardFileInputReportsAllMissingPaths(t *testing.T) {
+	dir := t.TempDir()
+	cmdutil.TestChdir(t, dir)
+	f, stdout, _, _ := cmdutil.TestFactory(t, docsCreateTestConfig(t, ""))
+
+	err := runDocsCreateShortcut(t, f, stdout, []string{
+		"+create",
+		"--api-version", "v2",
+		"--content", strings.Join([]string{
+			`<whiteboard type="svg" path="@missing.svg"></whiteboard>`,
+			`<whiteboard type="mermaid">@missing.mmd</whiteboard>`,
+			`<whiteboard type="plantuml" path="@missing.puml"></whiteboard>`,
+		}, "\n"),
+		"--as", "user",
+	})
+	if err == nil {
+		t.Fatal("expected aggregated whiteboard path error")
+	}
+	for _, want := range []string{
+		`whiteboard svg path "missing.svg" cannot be read`,
+		`whiteboard mermaid path "missing.mmd" cannot be read`,
+		`whiteboard plantuml path "missing.puml" cannot be read`,
+	} {
+		if !strings.Contains(err.Error(), want) {
+			t.Fatalf("error missing %q:\n%v", want, err)
+		}
+	}
+}
+
 func TestDocsCreateV2HTML5BlockRejectsInlineContent(t *testing.T) {
 	dir := t.TempDir()
 	cmdutil.TestChdir(t, dir)
--- a/skills/lark-doc/references/lark-doc-whiteboard.md
+++ b/skills/lark-doc/references/lark-doc-whiteboard.md
@@ -44,6 +44,8 @@ SubAgent 插入 SVG。
 </whiteboard>
 ```

+如果 Mermaid 已在本地文件中，可写成 `<whiteboard type="mermaid">@diagram.mmd</whiteboard>` 或 `<whiteboard type="mermaid" path="@diagram.mermaid"></whiteboard>`；CLI 会在写入前读取文件并展开为内联内容。
+
 ### 步骤 2B: SubAgent 使用 SVG 插入图表

 主 Agent 启动 SubAgent，让它用 `docs +create` / `docs +update` 插入：
@@ -56,6 +58,8 @@ SubAgent 插入 SVG。
 </whiteboard>
 ```

+如果 SVG 已在本地文件中，可写成 `<whiteboard type="svg" path="@diagram.svg"></whiteboard>`；PlantUML 文件同理使用 `<whiteboard type="plantuml" path="@sequence.puml"></whiteboard>` 或 `<whiteboard type="plantuml">@sequence.plantuml</whiteboard>`。
+
 Sub Agent 需要携带以下的最小上下文，以及后续的 [SVG 设计 Workflow] 章节指南：

 - doc token、插入位置（标题 / block_id / command）
--- a/skills/lark-doc/references/lark-doc-xml.md
+++ b/skills/lark-doc/references/lark-doc-xml.md
@@ -41,7 +41,7 @@ p, h1-h9, ul, ol, li, table, thead, tbody, tr, th, td, blockquote, pre, code, hr
 文档中可嵌入外部资源块（属于容器标签的特殊形式），需要额外语法创建：

 - `<img>` — `<img href="https://..."/>` 上传网络图片
- `<whiteboard>` — 简单图由 SubAgent 直接插入 `<whiteboard type="svg">完整自包含 SVG</whiteboard>`；复杂图使用 `<whiteboard type="blank"></whiteboard>` 先创建空白画板，再按 [`lark-doc-whiteboard.md`](lark-doc-whiteboard.md) 启动 SubAgent 调用 `lark-whiteboard` 写入；
+- `<whiteboard>` — 简单图由 SubAgent 直接插入 `<whiteboard type="svg">完整自包含 SVG</whiteboard>`；也可用本地文件简写 `<whiteboard type="svg" path="@diagram.svg"></whiteboard>`、`<whiteboard type="mermaid">@flow.mmd</whiteboard>`、`<whiteboard type="plantuml" path="@sequence.puml"></whiteboard>`，CLI 会写入前展开为内联内容；复杂图使用 `<whiteboard type="blank"></whiteboard>` 先创建空白画板，再按 [`lark-doc-whiteboard.md`](lark-doc-whiteboard.md) 启动 SubAgent 调用 `lark-whiteboard` 写入；
 - `<sheet>` — `<sheet type="blank"></sheet>` 空白；`<sheet sheet-id="SID" token="TOKEN"></sheet>` 复制已有
 - `<task>` — `<task task-id="GUID"></task>`，必传 task-id（任务 guid）
 - `<chat_card>` — `<chat_card chat-id="CHAT_ID"></chat_card>`，必传 chat-id