fix(search): preserve Unicode when truncating single results

docs-site/src/content/docs/guides/sidecars.md

@@ -283,3 +283,5 @@ Claude Code OAuth fingerprint precedent, but should be soak-tested with the
 intended account and workload.
 
 See the [Configuration reference](/reference/configuration/server/#sidecars) for every field.
+
+Long search queries and answers are shortened without splitting a UTF-16 surrogate pair, so a single character such as 😀 is never cut in half by the result limits. A sequence of several characters, such as emoji joined with a zero-width joiner, can still be cut between its parts.

docs-site/src/content/docs/ko/guides/sidecars.md

@@ -158,3 +158,5 @@ OpenCodex가 `web_search` 가로채기를 멈추고 Codex 통합이 `~/.codex/co
 파일을 직접 고치고 싶다면 이전처럼 `config.json`에서 `enabled`를 `false`로 두면 됩니다. Anthropic OAuth 검색과 이미지 설명은 기존 Claude Code OAuth
 fingerprint 선례를 따르지만, 실제 계정과 작업량으로 충분히 soak test하는 편이 좋습니다. 전체
 필드는 [설정 레퍼런스](/ko/reference/configuration/server/#sidecars)를 참고하세요.
+
+긴 검색어와 답변은 UTF-16 서로게이트 쌍을 나누지 않는 위치에서 줄이므로, 결과 길이 제한 때문에 😀 같은 한 글자가 반으로 잘리지 않습니다. 다만 제로 폭 결합자(ZWJ)로 이어진 이모지처럼 여러 글자로 이루어진 조합은 중간에서 잘릴 수 있습니다.

scripts/test-layout/layout-adapters.json

@@ -1,4 +1,5 @@
 {
+  "fix-search-unicode-clamp.test.ts": "web-search",
   "cursor-native-grep-globs.test.ts": "providers/cursor",
   "tool-identity-hash.test.ts": "adapters",
   "eventstream-fragmentation.test.ts": "responses",

src/web-search/format-result.ts

@@ -10,8 +10,9 @@ const MAX_TOTAL_CHARS = 8000;
 /** Bound for a batched error status so every query's status line stays small. */
 const MAX_BATCH_ERROR_CHARS = 300;
 
+/** Truncate to at most `max` UTF-16 units without splitting a surrogate pair, then mark the cut. */
 function clamp(s: string, max: number): string {
-  return s.length <= max ? s : `${s.slice(0, max)}\n…[truncated]`;
+  return s.length <= max ? s : `${safePrefix(s, max)}\n…[truncated]`;
 }
 
 /** A prefix of at most `max` UTF-16 units that never ends on half of a surrogate pair. */
@@ -25,7 +26,7 @@ function safePrefix(s: string, max: number): string {
 
 /** Clamp and sanitize a query string for display in the tool result boundary. */
 function safeQuery(q: string): string {
-  const clamped = q.length <= 200 ? q : `${q.slice(0, 200)}…`;
+  const clamped = q.length <= 200 ? q : `${safePrefix(q, 200)}…`;
   // Strip angle brackets so the query can't close/open XML-ish boundary tags.
   return clamped.replace(/[<>]/g, "");
 }

structure/data-planes/search.md

@@ -57,3 +57,5 @@ routing. Do not infer an exact account from caller-supplied account headers, pro
 selection, connection identity, or other ambient state; concurrent threads could cross-route
 credentials. Extending exact routing to those endpoints requires an opaque client correlation that
 can be bound server-side to a previously validated selector.
+
+Single-result query and answer truncation in `src/web-search/format-result.ts` uses the same surrogate-safe prefix as batched results. The existing UTF-16 limits and truncation markers remain unchanged.

tests/fixtures/test-layout-expected.json

@@ -1,4 +1,6 @@
-{ "provider-input-modalities.test.ts": "providers", "turn-membership.test.ts": "responses", "turn-membership-ingress.test.ts": "responses", "workbuddy-quota.test.ts": "providers", "workbuddy-login.test.ts": "oauth", "workbuddy-adapter.test.ts": "adapters", "workbuddy-provider.test.ts": "providers", "link-tunnel-history.test.ts": "clients", "link-readiness.test.ts": "clients", "link-diagnostics.test.ts": "clients", "combo-description.test.ts": "routing", "combo-description-catalog.test.ts": "codex-integration", "catalog-auto-review-override.test.ts": "codex-integration", "management-auto-review-settings.test.ts": "server", "management-auto-review-auth.test.ts": "server", "prefer-model-alias-ids.test.ts": "server", "responses-mimo-envelope.test.ts": "responses", "responses-call-id-linkage.test.ts": "responses", "responses-phase-inference.test.ts": "responses",
+{
+  "fix-search-unicode-clamp.test.ts": "web-search",
+  "provider-input-modalities.test.ts": "providers", "turn-membership.test.ts": "responses", "turn-membership-ingress.test.ts": "responses", "workbuddy-quota.test.ts": "providers", "workbuddy-login.test.ts": "oauth", "workbuddy-adapter.test.ts": "adapters", "workbuddy-provider.test.ts": "providers", "link-tunnel-history.test.ts": "clients", "link-readiness.test.ts": "clients", "link-diagnostics.test.ts": "clients", "combo-description.test.ts": "routing", "combo-description-catalog.test.ts": "codex-integration", "catalog-auto-review-override.test.ts": "codex-integration", "management-auto-review-settings.test.ts": "server", "management-auto-review-auth.test.ts": "server", "prefer-model-alias-ids.test.ts": "server", "responses-mimo-envelope.test.ts": "responses", "responses-call-id-linkage.test.ts": "responses", "responses-phase-inference.test.ts": "responses",
     "claude-intercept-acl-barrier.test.ts": "claude-integration",
     "codex-catalog-heal-reanchor.test.ts": "codex-integration",
     "tool-identity-hash.test.ts": "adapters",

tests/web-search/fix-search-unicode-clamp.test.ts

@@ -0,0 +1,27 @@
+import { describe, expect, test } from "bun:test";
+import { formatWebSearchResult } from "../../src/web-search/format-result";
+
+describe("single search result Unicode boundaries", () => {
+  test.each([false, true])("never splits query or answer surrogate pairs (structured=%s)", structured => {
+    const query = "q".repeat(199) + "😀tail";
+    const text = "a".repeat(3999) + "😀tail";
+    const result = formatWebSearchResult(query, { text, sources: [] }, structured);
+    const shown = structured ? JSON.parse(result.split("\n")[1]!) : { query: result, answer: result };
+    expect(shown.query).not.toMatch(/[\uD800-\uDBFF](?![\uDC00-\uDFFF])/u);
+    expect(shown.answer).not.toMatch(/[\uD800-\uDBFF](?![\uDC00-\uDFFF])/u);
+    expect(shown.answer).toContain("…[truncated]");
+  });
+  test("preserves complete pairs exactly at the limit", () => {
+    const text = "a".repeat(3998) + "😀";
+    const result = formatWebSearchResult("q".repeat(198) + "😀", { text, sources: [] }, true);
+    const data = JSON.parse(result.split("\n")[1]!);
+    expect(data.answer).toBe(text);
+    expect(data.query).toEndWith("😀");
+  });
+  test("keeps a pair that ends exactly at the limit when truncation occurs", () => {
+    const result = formatWebSearchResult("q".repeat(198) + "😀tail", { text: "a".repeat(3998) + "😀tail", sources: [] }, true);
+    const data = JSON.parse(result.split("\n")[1]!);
+    expect(data.answer).toBe("a".repeat(3998) + "😀\n…[truncated]");
+    expect(data.query).toBe("q".repeat(198) + "😀…");
+  });
+});