fix(chat): reconcile file-size baseline

Co-authored-by: diegosouzapw <8016841+diegosouzapw@users.noreply.github.com>

config/quality/file-size-baseline.json

@@ -1,5 +1,9 @@
 {
   "_rebaseline_2026_08_09_9296_adobe_media_capabilities": "PR #9296 (artickc, fix/adobe-firefly-model-capabilities) own growth: src/app/api/v1/models/catalog.ts 1590->1597 (+7). The image and video catalog serializers now expose the already-normalized Adobe Firefly discovery capability data (media_capabilities, plus the existing video modality/size fields) at their only response-emission chokepoints. The discovery parser and capability normalization remain in open-sse/services/adobeFireflyModels.ts; extracting these seven serialization fields would obscure the catalog contract. Covered by tests/unit/adobe-firefly.test.ts and tests/unit/image-upscale.test.ts.",
+  "_rebaseline_2026_08_08_v3850_base_drift_batch_9757": "Base drift on release/v3.8.50, not own growth: the 08-06..08-08 merge batches grew 12 already-frozen (or newly-landed) files without carrying their rebaselines — the dedicated rebaseline PR #9616 was closed as 'superseded' but its file-size entries never actually reached the base, and later merges (#8894 combos page, #9539 EditConnectionModal, #8895 models route, #9294/#9293 catalog, #9541 db/core, #8970 tokenHealthCheck, #8925 mcp schemas+server, #8890 accountFallback, #9467 chat.ts, #8931 openai-to-kiro, ProxyRegistryManager) kept growing them. All 12 values re-measured on THIS branch's tree (= pure tip + this PR's 1-line chat.ts fix, which adds zero lines). This PR's own source changes (chat.ts identifier restore, stream.ts format carve-out) do not grow any frozen file past these values.",
+  "_rebaseline_2026_08_08_migration_135_collision": "fix(db): resolve migration version 135 numbering collision — #9449's 135_connection_runtime_state.sql and #8908's 135_migrate_model_capability_max_token.sql both claimed version 135 (#9449 branched before #8908 merged and never got renumbered before landing on release/v3.8.50), which threw 'Migration version collision detected' the moment ANY code touched the database — a fresh install/deploy from this tip cannot even boot. Renumbered the later-landing file to 140 (next free slot) and added the matching isSchemaAlreadyApplied('140') retroactive guard, matching the established pattern already used for the prior 135/136 -> 137/138 renumber in the same file. Own growth: src/lib/db/migrationRunner.ts 1084->1094 (+10, the new case block) — irreducible, matches the existing per-case guard pattern exactly. Covered by tests/unit/migration-135-numbering-collision.test.ts (2/2), confirmed failing (reproducing the exact live crash) against the pre-fix colliding filenames, passing after.",
+  "_rebaseline_2026_08_08_9183_reasoning_cache_index_sync": "Extracted fix(responses-api): sync reasoning-cache write index with the fixed read side (from the originally-authored #9183) — chatCore.ts's write side cached every response under a hardcoded messageIndex:0, and translator/index.ts's plain-turn (non-tool-call) cache-key lookup ALSO still hardcoded messageIndex 0 at its call site (a second, previously-undiscovered instance of the same hardcoding bug, found while re-verifying this fix against the current upstream tip — the two never agreed once a conversation went past its first assistant turn, so DeepSeek/Xiaomi-mimo plain-turn reasoning replay silently missed the cache). Own growth: open-sse/handlers/chatCore.ts 5034->5042 (+8, computing messageIndex from the incoming request's message count at both the streaming and non-streaming cache-write call sites) — irreducible call-site wiring. Covered by tests/unit/reasoning-cache.test.ts (new end-to-end write/read regression test, rebaselined below) and tests/unit/translator-helper-branches.test.ts fixture updates. Other #9183 sub-fixes (output_index collision prevention, reasoning-content-alias generalization) were originally assumed already superseded by upstream's own independent fix — a live incident 2026-08-08 disproved that for the message-vs-tool-call collision case specifically (fixed separately in #9822); not re-extracted here since this PR's own scope is the narrower messageIndex sync only.",
+  "_rebaseline_2026_08_02_9259_rolling_rpm": "PR #9259 (issue #8733) own growth: open-sse/services/rateLimitManager.ts baseline 1060->1167 (+107; final source 1153). The existing withRateLimit chokepoint now composes process-local rolling RPM leases with Bottleneck admission, releases pre-dispatch leases on queue timeout/abort/connection disable, preserves caller abort reasons, and wires 429/header state into the extracted rollingRpmGate.ts. The remaining growth is irreducible lifecycle wiring at the dispatch boundary plus the real watchdog test hooks needed to verify queued-wedge recovery; moving it further would obscure lease ownership and Bottleneck cleanup. Covered by the focused rate-limit manager/sliding-window suite (33/33); distributed multi-instance coordination remains explicitly out of scope.",
   "_rebaseline_2026_07_24_8470_hyperagent_sticky_thread": "PR #8470 (artickc, fix/hyperagent-tool-loop-thread-sticky) own growth: open-sse/executors/hyperagent.ts 936->1025 (wc -l; check-file-size.mjs counts via split(\"\\n\").length so the gate sees 937->1026, +89, crosses the 1000 cap). Fixes a real bug where a reverse-conversion proxy (text-Intent/JSON to Claude Code native tool_calls) rewrites assistant messages between agentic tool-loop turns, breaking HyperAgent’s conversation-prefix fingerprint and cold-starting the thread mid tool-loop. Adds Anthropic tool_use/tool_result flattening to extractMessageText() plus a new rootUserFingerprint()/root-key lookup tier in resolveHyperAgentThreadBinding()/storeHyperAgentThreadAfterTurn() so the thread stays sticky across the tool loop. Cohesive additions inside the existing single-file executor; not extractable without splitting the executor mid-request-flow. Covered by tests/unit/executor-hyperagent.test.ts (19/19, +5 new cases for tool_result/tool_use flattening + root-key stickiness). Pre-merge review flagged a cross-conversation root-key collision risk (tracked in the PR’s own mandatory pre-merge checklist, not yet addressed) — unrelated to this file-size ratchet, tracked separately by /fix-prs.",
   "_rebaseline_2026_07_25_8494_capability_filter_fail_closed": "PR #8494 (fix/capability-filters-fail-closed, #8488) own growth: open-sse/services/combo.ts 3640->3693 (+53) adds a fail-closed guard after filterTargetsByRequestCompatibility() — when every eligible target is excluded by request-capability filtering (vision/tools/etc) instead of quota/health, the combo now returns an explicit `capability_mismatch` 400 (describeCapabilityFilterExhaustion, imported from combo/comboStructure.ts) rather than silently falling through to a generic no-targets error, plus a `compatFilterFailOpen` escape hatch (combo config OR settings) mirrored at both the main/auto and round-robin call sites for symmetry. combo/comboStructure.ts (previously under cap, un-frozen) grows 794->918 (+124) — new home for describeCapabilityFilterExhaustion + providerSupportsEmulatedToolCalling (#5240 emulated tool-calling exemption so fail-closed does not regress prompt-emulation-only combos like all-chatgpt-web). Irreducible orchestration wiring at the existing filter chokepoint (same precedent as #7301's universal-cooldown-retry generalization). Companion test tests/unit/combo-routing-engine.test.ts 3409->3449 (+40, fail-closed/fail-open coverage across both call sites) also rebaselined. Covered by tests/unit/8488-capability-filter-fail-closed.test.ts (new) + 95/95 passing across both files. Structural shrink of combo.ts tracked in #3501.",
   "_rebaseline_2026_07_25_8499_ts7_result_union_predicates": "PR #8499 (backryun, chore/ts7-types-executor-scattered) own growth: muse-spark-web.ts 1396->1405 (+9, irreducible). Under this workspace's `strictNullChecks: false`, the boolean-literal discriminant on `GraphqlResult` (`{ ok: true } | { ok: false; error: string }`) narrows the positive `.ok===true` branch but leaves `!result.ok` at the full union under TS7, making `.error` unreachable to the checker at the two call sites (warmup, mode-switch). Fixed by adding a single `isGraphqlFailure()` type-predicate helper (doc comment + 3-line body) reused at both call sites instead of duplicating the predicate inline — not extractable to a shared module without splitting a single-file executor's local narrowing helper out of its own file. Covered by the existing muse-spark-web executor test suite (no behavior change, pure narrowing fix).",
@@ -162,6 +166,7 @@
   "cap": 1000,
   "testCap": 1000,
   "testFrozen": {
+    "tests/unit/reasoning-cache.test.ts": 1035,
     "_rebaseline_2026_06_27_5193_antigravity_test": "#5193 own test growth: oauth-providers-config.test.ts 870->873 (+3: antigravity projectId assertion + 50ms tick for the now fire-and-forget onboarding, matching the no-PKCE/no-openid flow).",
     "_rebaseline_2026_07_02_5928_base_red": "web-cookie-providers-new.test.ts 845->850: #5928 (test(security) Kimi Web URL host parse, CodeQL #689) grew the file +5 lines and merged into release/v3.8.44 WITHOUT rebaselining, leaving a fast-gates base-red that blocked every subsequent PR->release. Test growth is legitimate (a security regression test); maintainer absorbs the drift here. Frozen at 850.",
     "_rebaseline_2026_07_09_6126_clinepass_dualauth": "#6126 (ClinePass dual-auth) own test growth: oauth-providers-config.test.ts 842->845 (+3: clinepass key/config/required-fields entries reusing the Cline WorkOS flow config, needed after registering clinepass in the oauth.ts PROVIDERS enum).",
@@ -350,7 +355,7 @@
     "open-sse/executors/deepseek-web.ts": 1148,
     "open-sse/executors/grok-web.ts": 1044,
     "open-sse/executors/muse-spark-web.ts": 1405,
-    "open-sse/handlers/chatCore.ts": 5034,
+    "open-sse/handlers/chatCore.ts": 5042,
     "open-sse/handlers/imageGeneration.ts": 3101,
     "open-sse/handlers/responseSanitizer.ts": 1128,
     "open-sse/handlers/search.ts": 1536,

open-sse/handlers/chatCore.ts

@@ -207,7 +207,6 @@ import { stageTrace } from "./chatCore/stageTrace.ts";
 import { attachCompressionUsageReceiptAfterAnalytics as attachCompressionUsageReceiptAfterAnalyticsFor } from "./chatCore/compressionUsageReceipt.ts";
 import { prepareUpstreamBody } from "./chatCore/upstreamBody.ts";
 import { getQuotaScopeLabelForProvider } from "../services/antigravityQuotaFamily.ts";
-
 import {
   getCallLogPipelineCaptureStreamChunks,
   getCallLogPipelineMaxSizeBytes,
@@ -367,9 +366,7 @@ import {
   isTpmExhausted,
   isRpmExhausted,
 } from "../services/geminiRateLimitTracker.ts";
-
 import { isSmallEnoughForSemanticCache } from "../utils/estimateSize.ts";
-
 /**
  * Core chat handler - shared between SSE and Worker
  * Returns { success, response, status, error } for caller to handle fallback
@@ -389,10 +386,8 @@ import { isSmallEnoughForSemanticCache } from "../utils/estimateSize.ts";
  * @param {boolean} options.isCombo - Whether this request is from a combo
  * @param {string} options.connectionId - Connection ID for settings lookup
  */
-
 // extractSystemRoleMessages extracted to chatCore/claudeSystemRole.ts (#3501); re-exported above so
 // existing importers (e.g. tests/unit/system-role-extraction.test.ts) keep resolving it from here.
-
 export async function handleChatCore({
   body,
   modelInfo,
@@ -428,7 +423,6 @@ export async function handleChatCore({
       /* fail open */
     }
   }
-
   // Per-request model-routing metadata (first extracted slice of the request-setup phase).
   const { apiFormat, customModelTargetFormat, requestedModel } = resolveChatCoreRequestSetup(
     modelInfo,
@@ -442,7 +436,6 @@ export async function handleChatCore({
   // (not Math.random) purely to satisfy CodeQL js/insecure-randomness — this id
   // is a log-correlation token, not a security secret.
   const traceId = globalThis.crypto.randomUUID().slice(0, 6);
-
   // Emit request.started event for real-time dashboard
   setImmediate(() => {
     emit("request.started", {
@@ -4333,9 +4326,14 @@ export async function handleChatCore({
     try {
       const firstChoice = translatedResponse?.choices?.[0];
       const msg = firstChoice?.message;
+      // The response being cached now will be replayed as history on the *next*
+      // turn, where the read side (translator/index.ts) keys the lookup by the
+      // message's real position in that future `messages` array — i.e. right
+      // after everything the client sent this turn.
+      const bodyMessages = (body as { messages?: unknown[] } | null | undefined)?.messages;
       cacheReasoningFromAssistantMessage(msg, provider, model, {
         requestId: skillRequestId,
-        messageIndex: 0,
+        messageIndex: Array.isArray(bodyMessages) ? bodyMessages.length : 0,
       });
     } catch {
       // Cache capture is non-critical — never block the response
@@ -4760,12 +4758,15 @@ export async function handleChatCore({
     // with tool_calls so it can be replayed on subsequent turns (DeepSeek V4, Kimi K2, etc.)
     if (normalizedStreamStatus === 200 && streamResponseBody) {
       try {
-        const body = streamResponseBody as Record<string, unknown>;
-        const choices = body.choices as { message?: Record<string, unknown> }[] | undefined;
+        const streamBody = streamResponseBody as Record<string, unknown>;
+        const choices = streamBody.choices as { message?: Record<string, unknown> }[] | undefined;
         const msg = choices?.[0]?.message;
+        // See the non-streaming capture above: messageIndex must match the
+        // position this message will occupy in the *next* turn's history.
+        const bodyMessages = (body as { messages?: unknown[] } | null | undefined)?.messages;
         cacheReasoningFromAssistantMessage(msg, provider, model, {
           requestId: skillRequestId,
-          messageIndex: 0,
+          messageIndex: Array.isArray(bodyMessages) ? bodyMessages.length : 0,
         });
       } catch {
         // Cache capture is non-critical — never block the stream
@@ -5025,7 +5026,6 @@ export async function handleChatCore({
     }),
   };
 }
-
 export function isTokenExpiringSoon(expiresAt, bufferMs = 5 * 60 * 1000) {
   if (!expiresAt) return false;
   const expiresAtMs = new Date(expiresAt).getTime();

open-sse/translator/index.ts

@@ -570,7 +570,7 @@ export function translateRequest(
 
       const cacheKey = hasToolCalls
         ? msg.tool_calls[0]?.id
-        : getAssistantMessageCacheKey(result, 0);
+        : getAssistantMessageCacheKey(result, messageIndex);
       if (cacheKey) {
         const cached = lookupReasoning(cacheKey);
         if (cached) {

tests/unit/reasoning-cache.test.ts

@@ -865,11 +865,12 @@ describe("Reasoning Replay Cache — Translator Replay", () => {
         }),
       },
     });
-    // NOTE: the non-tool-call cache key is built as `getAssistantMessageCacheKey(result, 0)`
-    // — the message index is hardcoded to 0 in the translator, so the key is always
-    // `request:<id>:message:0` regardless of the assistant message's actual position.
+    // The non-tool-call cache key is built as `getAssistantMessageCacheKey(result, messageIndex)`
+    // where messageIndex is the assistant message's real position in the `messages`
+    // array (index 1 here: user, assistant, user) — matching what the write side
+    // (chatCore.ts) now caches under once the response is generated.
     cacheReasoning(
-      "request:req-plain-1:message:0",
+      "request:req-plain-1:message:1",
       "deepseek",
       "deepseek-v4-pro",
       "Real cached plain-turn reasoning"
@@ -899,6 +900,60 @@ describe("Reasoning Replay Cache — Translator Replay", () => {
     );
     assert.equal(getReasoningCacheServiceStats().replays, 1);
   });
+
+  it("write side (chatCore's messageIndex) and read side (translateRequest) agree on the same key end-to-end", () => {
+    // Regression for a mismatch where chatCore.ts always cached under
+    // `messageIndex: 0` (the position of the response within *its own* choices
+    // array) while translateRequest's read side looked up the message's real
+    // position in the *next* turn's full history — the two never agreed once a
+    // conversation went past its first assistant turn, so replay silently
+    // fell back to the placeholder in real multi-turn usage.
+    clearReasoningCacheAll();
+    clearModelsDevCapabilities();
+    saveModelsDevCapabilities({
+      deepseek: {
+        "deepseek-v4-pro": buildCapability({
+          interleaved_field: "reasoning_content",
+          reasoning: true,
+          tool_call: true,
+        }),
+      },
+    });
+
+    // Turn 1: the incoming request has a single user message (length 1), so
+    // the assistant response chatCore is about to cache will occupy index 1
+    // once it's appended to history for turn 2 — mirroring
+    // `messageIndex: bodyMessages.length` in chatCore.ts.
+    const turn1RequestBody = { messages: [{ role: "user", content: "hi" }] };
+    cacheReasoningFromAssistantMessage(
+      { role: "assistant", content: "Hello! How can I help?", reasoning_content: "real reasoning" },
+      "deepseek",
+      "deepseek-v4-pro",
+      { requestId: "req-e2e-1", messageIndex: turn1RequestBody.messages.length }
+    );
+
+    // Turn 2: client replays the full history including the cached assistant
+    // turn, now genuinely at index 1.
+    const translated = translateRequest(
+      FORMATS.OPENAI,
+      FORMATS.OPENAI,
+      "deepseek-v4-pro",
+      {
+        request_id: "req-e2e-1",
+        messages: [
+          { role: "user", content: "hi" },
+          { role: "assistant", content: "Hello! How can I help?" },
+          { role: "user", content: "tell me more" },
+        ],
+      },
+      false,
+      null,
+      "deepseek"
+    );
+
+    assert.equal(translated.messages[1].reasoning_content, "real reasoning");
+    assert.equal(getReasoningCacheServiceStats().replays, 1);
+  });
 });
 
 describe("Reasoning Replay Cache — API Route", () => {

tests/unit/translator-helper-branches.test.ts

@@ -632,7 +632,7 @@ test("translateRequest replays cached reasoning-only messages when interleaved f
     },
   });
   cacheReasoningByKey(
-    "request:req_reasoning_only:message:0",
+    "request:req_reasoning_only:message:1",
     "deepseek",
     "deepseek-v4-flash",
     "cached reasoning only"
@@ -690,138 +690,131 @@ test("translateRequest does not replay reasoning-only messages for non-DeepSeek
   clearReasoningCacheAll();
 });
 
-  test("translateRequest uses Kimi Coding's empty thinking marker instead of cached replay", () => {
-    clearReasoningCacheAll();
-    cacheReasoningByKey(
-      "toolu_kimi_claude",
-      "kimi-coding",
-      "kimi-for-coding",
-      "cached thinking for Kimi tool call"
-    );
+test("translateRequest uses Kimi Coding's empty thinking marker instead of cached replay", () => {
+  clearReasoningCacheAll();
+  cacheReasoningByKey(
+    "toolu_kimi_claude",
+    "kimi-coding",
+    "kimi-for-coding",
+    "cached thinking for Kimi tool call"
+  );
 
-    // Claude-format request: assistant has tool_use in content[] but NO thinking block
-    // This simulates the scenario that causes infinite loops
-    const result = translateRequest(
-      FORMATS.OPENAI,
-      FORMATS.CLAUDE,
-      "kimi-for-coding",
-      {
-        reasoning_effort: "high",
-        messages: [
-          { role: "user", content: "read the file" },
-          {
-            role: "assistant",
-            content: [
-              {
-                type: "tool_use",
-                id: "toolu_kimi_claude",
-                name: "read_file",
-                input: { path: "test.ts" },
-              },
-            ],
-          },
-          { role: "tool", tool_call_id: "toolu_kimi_claude", content: "file data" },
-        ],
-      },
-      false,
-      null,
-      "kimi-coding"
-    );
+  // Claude-format request: assistant has tool_use in content[] but NO thinking block
+  // This simulates the scenario that causes infinite loops
+  const result = translateRequest(
+    FORMATS.OPENAI,
+    FORMATS.CLAUDE,
+    "kimi-for-coding",
+    {
+      reasoning_effort: "high",
+      messages: [
+        { role: "user", content: "read the file" },
+        {
+          role: "assistant",
+          content: [
+            {
+              type: "tool_use",
+              id: "toolu_kimi_claude",
+              name: "read_file",
+              input: { path: "test.ts" },
+            },
+          ],
+        },
+        { role: "tool", tool_call_id: "toolu_kimi_claude", content: "file data" },
+      ],
+    },
+    false,
+    null,
+    "kimi-coding"
+  );
 
-    const assistantMsg = result.messages.find((m) => m.role === "assistant");
-    assert.ok(assistantMsg, "assistant message should exist");
-    assert.ok(Array.isArray(assistantMsg.content), "content should be array");
+  const assistantMsg = result.messages.find((m) => m.role === "assistant");
+  assert.ok(assistantMsg, "assistant message should exist");
+  assert.ok(Array.isArray(assistantMsg.content), "content should be array");
 
-    // Kimi Code CLI 0.26 sends an explicit empty thinking marker before tool_use.
-    const thinkingBlock = assistantMsg.content.find((b) => b?.type === "thinking");
-    assert.ok(thinkingBlock, "thinking block should be injected");
-    assert.equal(thinkingBlock.thinking, "");
+  // Kimi Code CLI 0.26 sends an explicit empty thinking marker before tool_use.
+  const thinkingBlock = assistantMsg.content.find((b) => b?.type === "thinking");
+  assert.ok(thinkingBlock, "thinking block should be injected");
+  assert.equal(thinkingBlock.thinking, "");
 
-    // Thinking block should appear before tool_use
-    const thinkingIdx = assistantMsg.content.indexOf(thinkingBlock);
-    const toolUseIdx = assistantMsg.content.findIndex((b) => b?.type === "tool_use");
-    assert.ok(thinkingIdx < toolUseIdx, "thinking block should be before tool_use");
+  // Thinking block should appear before tool_use
+  const thinkingIdx = assistantMsg.content.indexOf(thinkingBlock);
+  const toolUseIdx = assistantMsg.content.findIndex((b) => b?.type === "tool_use");
+  assert.ok(thinkingIdx < toolUseIdx, "thinking block should be before tool_use");
 
-    assert.equal(getReasoningCacheServiceStats().replays, 0);
-    clearReasoningCacheAll();
-  });
+  assert.equal(getReasoningCacheServiceStats().replays, 0);
+  clearReasoningCacheAll();
+});
 
-  test("translateRequest uses an empty Kimi Coding thinking marker on cache miss", () => {
-    clearReasoningCacheAll();
+test("translateRequest uses an empty Kimi Coding thinking marker on cache miss", () => {
+  clearReasoningCacheAll();
 
-    const result = translateRequest(
-      FORMATS.OPENAI,
-      FORMATS.CLAUDE,
-      "kimi-for-coding",
-      {
-        reasoning_effort: "high",
-        messages: [
-          { role: "user", content: "do it" },
-          {
-            role: "assistant",
-            content: [
-              { type: "tool_use", id: "toolu_miss", name: "bash", input: { command: "ls" } },
-            ],
-          },
-          { role: "tool", tool_call_id: "toolu_miss", content: "output" },
-        ],
-      },
-      false,
-      null,
-      "kimi-coding"
-    );
+  const result = translateRequest(
+    FORMATS.OPENAI,
+    FORMATS.CLAUDE,
+    "kimi-for-coding",
+    {
+      reasoning_effort: "high",
+      messages: [
+        { role: "user", content: "do it" },
+        {
+          role: "assistant",
+          content: [{ type: "tool_use", id: "toolu_miss", name: "bash", input: { command: "ls" } }],
+        },
+        { role: "tool", tool_call_id: "toolu_miss", content: "output" },
+      ],
+    },
+    false,
+    null,
+    "kimi-coding"
+  );
 
-    const assistantMsg = result.messages.find((m) => m.role === "assistant");
-    assert.ok(assistantMsg, "assistant message should exist");
+  const assistantMsg = result.messages.find((m) => m.role === "assistant");
+  assert.ok(assistantMsg, "assistant message should exist");
 
-    const thinkingBlock =
-      Array.isArray(assistantMsg.content) &&
-      assistantMsg.content.find((b) => b?.type === "thinking");
-    assert.ok(thinkingBlock, "thinking block should be injected on cache miss");
-    assert.equal(thinkingBlock.thinking, "");
+  const thinkingBlock =
+    Array.isArray(assistantMsg.content) && assistantMsg.content.find((b) => b?.type === "thinking");
+  assert.ok(thinkingBlock, "thinking block should be injected on cache miss");
+  assert.equal(thinkingBlock.thinking, "");
 
-    clearReasoningCacheAll();
-  });
+  clearReasoningCacheAll();
+});
 
-  test("translateRequest does NOT inject duplicate thinking for Claude-format messages with existing thinking block", () => {
-    clearReasoningCacheAll();
+test("translateRequest does NOT inject duplicate thinking for Claude-format messages with existing thinking block", () => {
+  clearReasoningCacheAll();
 
-    const result = translateRequest(
-      FORMATS.OPENAI,
-      FORMATS.CLAUDE,
-      "kimi-for-coding",
-      {
-        messages: [
-          { role: "user", content: "hi" },
-          {
-            role: "assistant",
-            content: [
-              { type: "thinking", thinking: "I already have this" },
-              { type: "tool_use", id: "toolu_existing", name: "read", input: {} },
-            ],
-          },
-          { role: "tool", tool_call_id: "toolu_existing", content: "data" },
-        ],
-      },
-      false,
-      null,
-      "kimi-coding"
-    );
+  const result = translateRequest(
+    FORMATS.OPENAI,
+    FORMATS.CLAUDE,
+    "kimi-for-coding",
+    {
+      messages: [
+        { role: "user", content: "hi" },
+        {
+          role: "assistant",
+          content: [
+            { type: "thinking", thinking: "I already have this" },
+            { type: "tool_use", id: "toolu_existing", name: "read", input: {} },
+          ],
+        },
+        { role: "tool", tool_call_id: "toolu_existing", content: "data" },
+      ],
+    },
+    false,
+    null,
+    "kimi-coding"
+  );
 
-    const assistantMsg = result.messages.find((m) => m.role === "assistant");
-    const thinkingBlocks =
-      Array.isArray(assistantMsg.content) &&
-      assistantMsg.content.filter((b) => b?.type === "thinking");
-    assert.equal(
-      thinkingBlocks?.length,
-      1,
-      "should have exactly one thinking block (no duplicate)"
-    );
-    assert.equal(
-      thinkingBlocks[0].thinking,
-      "I already have this",
-      "original thinking should be preserved"
-    );
+  const assistantMsg = result.messages.find((m) => m.role === "assistant");
+  const thinkingBlocks =
+    Array.isArray(assistantMsg.content) &&
+    assistantMsg.content.filter((b) => b?.type === "thinking");
+  assert.equal(thinkingBlocks?.length, 1, "should have exactly one thinking block (no duplicate)");
+  assert.equal(
+    thinkingBlocks[0].thinking,
+    "I already have this",
+    "original thinking should be preserved"
+  );
 
-    clearReasoningCacheAll();
-  });
+  clearReasoningCacheAll();
+});