From 3fba807abf68adb507fcd0aedfc843ebf16041e5 Mon Sep 17 00:00:00 2001
From: Patrick Wozniak <email@patwoz.de>
Date: Thu, 28 May 2026 22:05:01 +0200
Subject: [PATCH] fix(stream): close thinking block before text or tool output

If upstream omits reasoning-end and starts emitting text-delta or
tool-call events, the thinking block remained open causing
thinking_start -> thinking_delta -> text_start -> ... -> thinking_end
ordering. Now endThinking() is called at the start of text-delta and
tool-call handlers so thinking_end always precedes text_start or
toolcall_start.

Adds 2 regression tests for the missing reasoning-end edge case.
---
 src/core.ts          |  2 ++
 tests/test-stream.ts | 60 ++++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 62 insertions(+)

diff --git a/src/core.ts b/src/core.ts
index f01198f..e80b44e 100644
--- a/src/core.ts
+++ b/src/core.ts
@@ -223,6 +223,7 @@ export function createStreamCommandCode(deps: CoreDependencies) {
 
         switch (event.type) {
           case "text-delta": {
+            endThinking()
             if (!textBlock) {
               textBlock = { type: "text", text: "" }
               output.content.push(textBlock)
@@ -286,6 +287,7 @@ export function createStreamCommandCode(deps: CoreDependencies) {
 
           case "tool-call": {
             endTextBlock()
+            endThinking()
             const toolCall: ToolCallContent = {
               type: "toolCall",
               id: stringValue(event.toolCallId) ?? "",
diff --git a/tests/test-stream.ts b/tests/test-stream.ts
index 800a972..57b630a 100644
--- a/tests/test-stream.ts
+++ b/tests/test-stream.ts
@@ -202,6 +202,66 @@ describe("streamCommandCode — successful streams", () => {
     if (done?.type !== "done") throw new Error("expected done")
     assert.equal(done.message.content[0]?.type, "thinking")
   })
+
+  it("closes thinking block before text when reasoning-end is missing", async () => {
+    server.mockResponse({
+      type: "success",
+      events: [
+        JSON.stringify({ type: "reasoning-start" }),
+        JSON.stringify({ type: "reasoning-delta", text: "thinking" }),
+        JSON.stringify({ type: "text-delta", text: "answer" }),
+        JSON.stringify({ type: "finish", finishReason: "stop" }),
+      ],
+    })
+    const { streamCommandCode } = createTestDeps({ apiBase: server.baseUrl() })
+
+    const events = await collectEvents(
+      streamCommandCode(makeModel(), makeContext(), { apiKey: "mock-key" }),
+    )
+
+    assert.deepEqual(eventTypes(events), [
+      "start",
+      "thinking_start",
+      "thinking_delta",
+      "thinking_end",
+      "text_start",
+      "text_delta",
+      "text_end",
+      "done",
+    ])
+  })
+
+  it("closes thinking block before tool-call when reasoning-end is missing", async () => {
+    server.mockResponse({
+      type: "success",
+      events: [
+        JSON.stringify({ type: "reasoning-start" }),
+        JSON.stringify({ type: "reasoning-delta", text: "thinking" }),
+        JSON.stringify({
+          type: "tool-call",
+          toolCallId: "call_1",
+          toolName: "read_file",
+          input: JSON.stringify({ path: "/tmp/x" }),
+        }),
+        JSON.stringify({ type: "finish", finishReason: "tool-calls" }),
+      ],
+    })
+    const { streamCommandCode } = createTestDeps({ apiBase: server.baseUrl() })
+
+    const events = await collectEvents(
+      streamCommandCode(makeModel(), makeContext(), { apiKey: "mock-key" }),
+    )
+
+    assert.deepEqual(eventTypes(events), [
+      "start",
+      "thinking_start",
+      "thinking_delta",
+      "thinking_end",
+      "toolcall_start",
+      "toolcall_end",
+      "done",
+    ])
+  })
 })
 
 describe("streamCommandCode — request serialization", () => {