From 870bfa94ed1dc991bae685ee1c4e528af14d72b0 Mon Sep 17 00:00:00 2001
From: Peter Steinberger <steipete@gmail.com>
Date: Thu, 22 Jan 2026 23:51:48 +0000
Subject: [PATCH] fix: skip tool id sanitization for openai responses

---
 CHANGELOG.md                                  |  1 +
 ...ed-runner.google-sanitize-thinking.test.ts | 14 ++++++++-----
 ...ed-runner.sanitize-session-history.test.ts | 21 ++++++++++++++++++-
 src/agents/pi-embedded-runner/google.ts       |  8 +------
 4 files changed, 31 insertions(+), 13 deletions(-)

diff --git a/CHANGELOG.md b/CHANGELOG.md
index 897fbd7fb..20aacf256 100644
--- a/CHANGELOG.md
+++ b/CHANGELOG.md
@@ -27,6 +27,7 @@ Docs: https://docs.clawd.bot
 - Auto-reply: only report a model switch when session state is available. (#1465) Thanks @robbyczgw-cla.
 - Control UI: resolve local avatar URLs with basePath across injection + identity RPC. (#1457) Thanks @dlauer.
 - Agents: surface concrete API error details instead of generic AI service errors.
+- Agents: avoid sanitizing tool call IDs for OpenAI responses to preserve Pi pairing.
 - Docs: fix gog auth services example to include docs scope. (#1454) Thanks @zerone0x.
 - macOS: prefer linked channels in gateway summary to avoid false “not linked” status.
 
diff --git a/src/agents/pi-embedded-runner.google-sanitize-thinking.test.ts b/src/agents/pi-embedded-runner.google-sanitize-thinking.test.ts
index 5e58e6d6c..ed4b52940 100644
--- a/src/agents/pi-embedded-runner.google-sanitize-thinking.test.ts
+++ b/src/agents/pi-embedded-runner.google-sanitize-thinking.test.ts
@@ -320,7 +320,7 @@ describe("sanitizeSessionHistory (google thinking)", () => {
     expect(assistant.content?.map((block) => block.type)).toEqual(["thinking"]);
   });
 
-  it("sanitizes tool call ids for OpenAI-compatible APIs", async () => {
+  it("sanitizes tool call ids for Google APIs", async () => {
     const sessionManager = SessionManager.inMemory();
     const longId = `call_${"a".repeat(60)}`;
     const input = [
@@ -338,17 +338,21 @@ describe("sanitizeSessionHistory (google thinking)", () => {
 
     const out = await sanitizeSessionHistory({
       messages: input,
-      modelApi: "openai-responses",
+      modelApi: "google-antigravity",
       sessionManager,
-      sessionId: "session:openai",
+      sessionId: "session:google",
     });
 
-    const assistant = out[0] as Extract<AgentMessage, { role: "assistant" }>;
+    const assistant = out.find(
+      (msg) => (msg as { role?: unknown }).role === "assistant",
+    ) as Extract<AgentMessage, { role: "assistant" }>;
     const toolCall = assistant.content?.[0] as { id?: string };
     expect(toolCall.id).toBeDefined();
     expect(toolCall.id?.length).toBeLessThanOrEqual(40);
 
-    const toolResult = out[1] as Extract<AgentMessage, { role: "toolResult" }>;
+    const toolResult = out.find(
+      (msg) => (msg as { role?: unknown }).role === "toolResult",
+    ) as Extract<AgentMessage, { role: "toolResult" }>;
     expect(toolResult.toolCallId).toBe(toolCall.id);
   });
 });
diff --git a/src/agents/pi-embedded-runner.sanitize-session-history.test.ts b/src/agents/pi-embedded-runner.sanitize-session-history.test.ts
index c7a9b3d05..7d812985b 100644
--- a/src/agents/pi-embedded-runner.sanitize-session-history.test.ts
+++ b/src/agents/pi-embedded-runner.sanitize-session-history.test.ts
@@ -73,7 +73,7 @@ describe("sanitizeSessionHistory", () => {
     );
   });
 
-  it("does not sanitize tool call ids for non-Google, non-OpenAI APIs", async () => {
+  it("does not sanitize tool call ids for non-Google APIs", async () => {
     vi.mocked(helpers.isGoogleModelApi).mockReturnValue(false);
 
     await sanitizeSessionHistory({
@@ -92,6 +92,25 @@ describe("sanitizeSessionHistory", () => {
     );
   });
 
+  it("does not sanitize tool call ids for openai-responses", async () => {
+    vi.mocked(helpers.isGoogleModelApi).mockReturnValue(false);
+
+    await sanitizeSessionHistory({
+      messages: mockMessages,
+      modelApi: "openai-responses",
+      provider: "openai",
+      sessionManager: mockSessionManager,
+      sessionId: "test-session",
+    });
+
+    expect(helpers.isGoogleModelApi).toHaveBeenCalledWith("openai-responses");
+    expect(helpers.sanitizeSessionMessagesImages).toHaveBeenCalledWith(
+      mockMessages,
+      "session:history",
+      expect.objectContaining({ sanitizeToolCallIds: false }),
+    );
+  });
+
   it("keeps reasoning-only assistant messages for openai-responses", async () => {
     vi.mocked(helpers.isGoogleModelApi).mockReturnValue(false);
 
diff --git a/src/agents/pi-embedded-runner/google.ts b/src/agents/pi-embedded-runner/google.ts
index 0ec334192..41f776568 100644
--- a/src/agents/pi-embedded-runner/google.ts
+++ b/src/agents/pi-embedded-runner/google.ts
@@ -40,12 +40,6 @@ const GOOGLE_SCHEMA_UNSUPPORTED_KEYWORDS = new Set([
   "minProperties",
   "maxProperties",
 ]);
-const OPENAI_TOOL_CALL_ID_APIS = new Set([
-  "openai",
-  "openai-completions",
-  "openai-responses",
-  "openai-codex-responses",
-]);
 const MISTRAL_MODEL_HINTS = [
   "mistral",
   "mixtral",
@@ -67,7 +61,7 @@ function isValidAntigravitySignature(value: unknown): value is string {
 
 function shouldSanitizeToolCallIds(modelApi?: string | null): boolean {
   if (!modelApi) return false;
-  return isGoogleModelApi(modelApi) || OPENAI_TOOL_CALL_ID_APIS.has(modelApi);
+  return isGoogleModelApi(modelApi);
 }
 
 function isMistralModel(params: { provider?: string | null; modelId?: string | null }): boolean {