From 1974ea94303bd14c144f83c50d17c35a0d9861cf Mon Sep 17 00:00:00 2001
From: luvs01 <luvs01@hanmail.net>
Date: Mon, 31 Aug 2026 10:44:53 +0900
Subject: [PATCH] fix(kiro): persist local terminal continuations

---
 src/server/responses/core.ts             | 22 +++++++---
 tests/server-kiro-completion-e2e.test.ts | 56 ++++++++++++++++++++++++
 2 files changed, 73 insertions(+), 5 deletions(-)

diff --git a/src/server/responses/core.ts b/src/server/responses/core.ts
index 1f56ed979a..158165ad17 100644
--- a/src/server/responses/core.ts
+++ b/src/server/responses/core.ts
@@ -5480,6 +5480,14 @@ async function handleResponsesInner(
           translatorBudget,
           ...(options.forceEmptyResponseId ? { responseId: "" } : {}),
           ...(options.onFirstOutput ? { onFirstOutput: options.onFirstOutput } : {}),
+          onCompletedResponse: (response: Record<string, unknown>) => {
+            rememberResponseState(
+              parsed._rawBody,
+              response,
+              continuationStateForResponse(),
+              responseStateOptions(adapterNeedsForcedContinuation(activeAdapter.name)),
+            );
+          },
         },
       );
       // Same lifetime tracking as every other streaming return in this function: the turn
@@ -5498,12 +5506,16 @@ async function handleResponsesInner(
         },
       );
     }
-    return new Response(
-      JSON.stringify(buildResponseJSON(terminalEvents, parsed._responseModelId ?? parsed.modelId, {
-        translatorBudget,
-      })),
-      { headers: { "Content-Type": "application/json" } },
+    const json = buildResponseJSON(terminalEvents, parsed._responseModelId ?? parsed.modelId, {
+      translatorBudget,
+    });
+    rememberResponseState(
+      parsed._rawBody,
+      json,
+      continuationStateForResponse(),
+      responseStateOptions(adapterNeedsForcedContinuation(activeAdapter.name)),
     );
+    return new Response(JSON.stringify(json), { headers: { "Content-Type": "application/json" } });
   }
   // One request-scoped transient-retry budget owner, declared here so BOTH the initial send
   // and the later recovery refetches (429, key/account rotation, OAuth replay) share it. A
diff --git a/tests/server-kiro-completion-e2e.test.ts b/tests/server-kiro-completion-e2e.test.ts
index 61b6181643..03b306074a 100644
--- a/tests/server-kiro-completion-e2e.test.ts
+++ b/tests/server-kiro-completion-e2e.test.ts
@@ -637,3 +637,59 @@ describe("Kiro local terminal accounting", () => {
     });
   }
 });
+
+describe("Kiro local terminal continuation", () => {
+  for (const stream of [true, false]) {
+    test(`a local terminal remains available through previous_response_id (stream=${stream})`, async () => {
+      const upstream = scriptedKiroUpstream([completionFrames("Yes — one JavaScript cell, many tool calls.")]);
+      saveConfig(kiroConfig(upstream.server.url.toString()));
+      const proxy = startServer(0);
+      try {
+        const terminal = await originalFetch(new URL("/v1/responses", proxy.url), {
+          method: "POST",
+          headers: { "content-type": "application/json" },
+          body: JSON.stringify({
+            model: "kiro-test/gpt-5.6-sol",
+            stream,
+            store: false,
+            input: [
+              { type: "message", role: "user", content: [{ type: "input_text", text: "what is code mode" }] },
+              {
+                type: "message",
+                role: "assistant",
+                phase: "final_answer",
+                content: [{ type: "output_text", text: "Code mode runs JavaScript that calls tools." }],
+              },
+            ],
+          }),
+        });
+        expect(terminal.status).toBe(200);
+        const terminalBody = await terminal.text();
+        const responseId = stream
+          ? responseEvents(terminalBody).find(event => event.name === "response.completed")?.data.response.id
+          : (JSON.parse(terminalBody) as { id: string }).id;
+        expect(responseId).toMatch(/^resp_/);
+        expect(upstream.requests).toHaveLength(0);
+
+        const followUp = await originalFetch(new URL("/v1/responses", proxy.url), {
+          method: "POST",
+          headers: { "content-type": "application/json" },
+          body: JSON.stringify({
+            model: "kiro-test/gpt-5.6-sol",
+            stream: false,
+            store: false,
+            previous_response_id: responseId,
+            input: "so it batches calls?",
+          }),
+        });
+
+        expect(followUp.status).toBe(200);
+        await followUp.text();
+        expect(upstream.requests).toHaveLength(1);
+      } finally {
+        await proxy.stop(true);
+        upstream.server.stop(true);
+      }
+    });
+  }
+});