Expand WebSearch JSON result wrappers

Author: SqlRush

docs/claude-code-go-rewrite-plan.md

@@ -186,6 +186,7 @@ test/parity/                 # golden tests against TS/official behavior
 - 本轮补充：WebFetch 文本 body 现在会按 BOM、`Content-Type` charset 或 HTML `<meta charset>`/`http-equiv` charset 解码常见网页编码，包括 UTF-8/UTF-16LE/UTF-16BE、Latin-1 和 Windows-1252，并在 structured content 暴露归一化 `charset`。
 - 本轮补充：WebSearch JSON parser 现在会递归解包 `web`、`response`、`search`、`hits`、`documents`、`records`、`entries` 等常见后端 wrapper，保留 URL 去重和 domain filter。
 - 本轮补充：WebSearch JSON result parser 现在支持 `pageUrl`/`targetUrl`/`source_url`/`formattedUrl` 等 URL aliases、`htmlTitle`/`htmlSnippet` 等 HTML 标记字段清理、嵌套 URL object，以及 `deepLinks`/`siteLinks` 子结果递归解析。
+- 本轮补充：WebSearch JSON parser 现在继续覆盖 `answerBox`/`answer_box`、`knowledgeGraph`/`knowledge_graph`、`news`/`news_results`、`topStories`、`peopleAlsoAsk` 和 `related_questions` 等常见搜索后端 wrapper，并识别 `website`/`sourceLink` URL alias、`question` title alias、`answer`/`excerpt` snippet alias。
 - 本轮补充：WebSearch `query` schema 现在按官方 `min(2)` 约束拒绝单字符查询，通用 tool schema validator 同步支持 `minLength`，让工具定义可直接表达字符串最小长度契约。
 - 本轮补充：WebSearch domain filters 现在在 schema 层声明 array `items:string`，通用 tool schema validator 同步支持 `items` 校验；`allowed_domains`/`blocked_domains` 会拒绝空字符串、URL/port、非法 wildcard 和非域名 label。
 - 本轮补充：通用 tool schema validator 现在支持 `enum`，可直接执行 Grep output mode、NotebookEdit edit mode/cell type、Todo status/priority、Task target/action、LSP severity 等工具 schema 的枚举契约。

docs/first-second-parity-audit.md

@@ -127,6 +127,7 @@ Covered behavior:
 - `WebFetch` text bodies now decode common web charsets from BOM, `Content-Type`, or HTML `<meta charset>`/`http-equiv` declarations, including UTF-8/UTF-16LE/UTF-16BE, Latin-1, and Windows-1252, and structured results expose the normalized `charset`.
 - `WebSearch` initial HTML/JSON-search adapter, including query/max-result/timeout/domain-filter validation, injectable search endpoint, DuckDuckGo HTML link parsing, DuckDuckGo subdomain redirect unwrapping, common JSON result shapes plus nested backend wrappers such as `web`/`response`/`hits`/`documents`, DuckDuckGo result snippet extraction, domain allow/block filtering, structured result payloads, and query-based permission-rule matching.
 - `WebSearch` JSON result parsing now accepts additional backend aliases such as `pageUrl`/`targetUrl`/`source_url`/`formattedUrl`, cleans HTML-bearing title/snippet fields like `htmlTitle`/`htmlSnippet`, unwraps nested URL objects, and recursively collects `deepLinks`/`siteLinks` child results.
+- `WebSearch` JSON result parsing now also unwraps answer/knowledge/news/people-also-ask style backend containers such as `answerBox`, `knowledgeGraph`, `news_results`, `topStories`, `peopleAlsoAsk`, and `related_questions`, while accepting `website`/`sourceLink`, `question`, `answer`, and `excerpt` result aliases.
 - `Grep` now accepts whole-word matching through `word_regexp`/`wordRegexp`/`word-regexp`/`-w`, applying word-boundary filtering for both regex and fixed-string patterns and preserving quoted boolean coercion.
 - `Grep` now accepts inverted matching through `invert_match`/`invertMatch`/`invert-match`/`-v`, applying the inverted line set consistently for files-with-matches, content, count, and multiline span output modes while preserving quoted boolean coercion.
 - `Grep` content output now accepts ripgrep-style no-line-number controls such as `--no-line-number` and `-N`, plus snake/camel adjacent aliases, to disable default line-number output with structured state preserved.

internal/tools/web/web_search.go

@@ -278,6 +278,9 @@ func collectJSONSearchResults(value any, base *url.URL, results *[]searchResult,
 			"results", "organic_results", "organicResults", "items", "value", "data",
 			"webPages", "web_pages", "web", "response", "search",
 			"hits", "documents", "records", "entries", "organic",
+			"answer_box", "answerBox", "knowledge_graph", "knowledgeGraph",
+			"news", "news_results", "newsResults", "top_stories", "topStories",
+			"people_also_ask", "peopleAlsoAsk", "related_questions", "relatedQuestions",
 			"deepLinks", "deep_links", "siteLinks", "sitelinks", "pages", "matches",
 		} {
 			if child, ok := typed[key]; ok {
@@ -295,7 +298,7 @@ func searchResultFromJSONObject(obj map[string]any, base *url.URL) (searchResult
 	}
 	title := cleanJSONSearchText(jsonStringField(obj,
 		"title", "name", "headline", "heading",
-		"htmlTitle", "html_title",
+		"question", "label", "htmlTitle", "html_title",
 	))
 	if strings.TrimSpace(title) == "" {
 		title = resolved
@@ -304,6 +307,7 @@ func searchResultFromJSONObject(obj map[string]any, base *url.URL) (searchResult
 		"snippet", "description", "content", "text",
 		"htmlSnippet", "html_snippet",
 		"summary", "extract", "abstract", "body", "caption",
+		"answer", "excerpt",
 	))
 	return searchResult{
 		Title:   title,
@@ -319,6 +323,8 @@ func jsonSearchURLField(obj map[string]any) string {
 		"targetUrl", "targetURL", "target_url",
 		"webUrl", "webURL", "web_url",
 		"sourceUrl", "sourceURL", "source_url",
+		"sourceLink", "source_link",
+		"website", "site",
 	); raw != "" {
 		return raw
 	}

internal/tools/web/web_search_test.go

@@ -229,6 +229,65 @@ func TestWebSearchParsesAlternateJSONFieldAliases(t *testing.T) {
 	}
 }
 
+func TestWebSearchParsesSearchBackendWrapperObjects(t *testing.T) {
+	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
+		w.Header().Set("Content-Type", "application/json")
+		_, _ = w.Write([]byte(`{
+			"answerBox": {
+				"title": "Answer Box",
+				"link": "https://example.com/answer",
+				"answer": "direct answer"
+			},
+			"knowledgeGraph": {
+				"title": "Knowledge Graph",
+				"website": "https://example.com/kg",
+				"description": "knowledge description"
+			},
+			"news": {
+				"results": [
+					{"title": "News Result", "url": "https://news.example.com/story", "excerpt": "news excerpt"}
+				]
+			},
+			"peopleAlsoAsk": [
+				{"question": "Question Result", "sourceLink": "https://example.com/question", "snippet": "question snippet"}
+			],
+			"related_questions": [
+				{"question": "Duplicate Question", "link": "https://example.com/question", "snippet": "duplicate"}
+			]
+		}`))
+	}))
+	defer server.Close()
+	executor := webExecutor(t)
+	result, err := executor.Execute(tool.Context{
+		Context: context.Background(),
+		Metadata: map[string]any{
+			MetadataWebSearchEndpointKey: server.URL,
+		},
+	}, contracts.ToolUse{
+		ID:    "toolu_search_backend_wrappers",
+		Name:  "WebSearch",
+		Input: json.RawMessage(`{"query":"backend wrappers","allowed_domains":["example.com"],"max_results":10}`),
+	}, nil)
+	if err != nil {
+		t.Fatal(err)
+	}
+	results, ok := result.StructuredContent["results"].([]map[string]any)
+	if !ok || len(results) != 4 {
+		t.Fatalf("structured results = %#v", result.StructuredContent["results"])
+	}
+	wantTitles := []string{"Answer Box", "Knowledge Graph", "News Result", "Question Result"}
+	wantSnippets := []string{"direct answer", "knowledge description", "news excerpt", "question snippet"}
+	for i := range wantTitles {
+		if results[i]["title"] != wantTitles[i] || results[i]["snippet"] != wantSnippets[i] {
+			t.Fatalf("result %d = %#v", i, results[i])
+		}
+	}
+	content := result.Content.(string)
+	if !strings.Contains(content, "Knowledge Graph") || !strings.Contains(content, "news excerpt") || strings.Contains(content, "Duplicate Question") {
+		t.Fatalf("content = %#v", content)
+	}
+}
+
 func TestWebSearchBlockedDomainsAndNoResults(t *testing.T) {
 	server := httptest.NewServer(http.HandlerFunc(func(w http.ResponseWriter, r *http.Request) {
 		w.Header().Set("Content-Type", "text/html")