Deduplicate repeated tool call IDs for OpenAI-compatible APIs (#40996)

Merged via squash.

Prepared head SHA: 38d8048359
Co-authored-by: xaeon2026 <264572156+xaeon2026@users.noreply.github.com>
Co-authored-by: frankekn <4488090+frankekn@users.noreply.github.com>
Reviewed-by: @frankekn

CHANGELOG.md

@@ -32,6 +32,7 @@ Docs: https://docs.openclaw.ai
 - Models/openai-completions: default non-native OpenAI-compatible providers to omit tool-definition `strict` fields unless users explicitly opt back in, so tool calling keeps working on providers that reject that option. (#45497) Thanks @sahancava.
 - WhatsApp/reconnect: restore the append recency filter in the extension inbox monitor and handle protobuf `Long` timestamps correctly, so fresh post-reconnect append messages are processed while stale history sync stays suppressed. (#42588) thanks @MonkeyLeeT.
 - WhatsApp/login: wait for pending creds writes before reopening after Baileys `515` pairing restarts in both QR login and `channels login` flows, and keep the restart coverage pinned to the real wrapped error shape plus per-account creds queues. (#27910) Thanks @asyncjason.
+- Agents/openai-compatible tool calls: deduplicate repeated tool call ids across live assistant messages and replayed history so OpenAI-compatible backends no longer reject duplicate `tool_call_id` values with HTTP 400. (#40996) Thanks @xaeon2026.
 
 ### Fixes
 

src/agents/pi-embedded-runner.sanitize-session-history.test.ts

@@ -2,6 +2,7 @@ import type { AgentMessage } from "@mariozechner/pi-agent-core";
 import type { AssistantMessage, UserMessage, Usage } from "@mariozechner/pi-ai";
 import { beforeEach, describe, expect, it, vi } from "vitest";
 import {
+  expectOpenAIResponsesStrictSanitizeCall,
   loadSanitizeSessionHistoryWithCleanMocks,
   makeMockSessionManager,
   makeInMemorySessionManager,
@@ -247,7 +248,24 @@ describe("sanitizeSessionHistory", () => {
     expect(result).toEqual(mockMessages);
   });
 
-  it("passes simple user-only history through for openai-completions", async () => {
+  it("sanitizes tool call ids for OpenAI-compatible responses providers", async () => {
+    setNonGoogleModelApi();
+
+    await sanitizeSessionHistory({
+      messages: mockMessages,
+      modelApi: "openai-responses",
+      provider: "custom",
+      sessionManager: mockSessionManager,
+      sessionId: TEST_SESSION_ID,
+    });
+
+    expectOpenAIResponsesStrictSanitizeCall(
+      mockedHelpers.sanitizeSessionMessagesImages,
+      mockMessages,
+    );
+  });
+
+  it("sanitizes tool call ids for openai-completions", async () => {
     setNonGoogleModelApi();
 
     const result = await sanitizeSessionHistory({

src/agents/pi-embedded-runner/run/attempt.test.ts

@@ -702,6 +702,26 @@ describe("wrapStreamFnTrimToolCallNames", () => {
     expect(finalToolCall.name).toBe("read");
     expect(finalToolCall.id).toBe("call_42");
   });
+
+  it("reassigns duplicate tool call ids within a message to unique fallbacks", async () => {
+    const finalToolCallA = { type: "toolCall", name: " read ", id: "  edit:22  " };
+    const finalToolCallB = { type: "toolCall", name: " write ", id: "edit:22" };
+    const finalMessage = { role: "assistant", content: [finalToolCallA, finalToolCallB] };
+    const baseFn = vi.fn(() =>
+      createFakeStream({
+        events: [],
+        resultMessage: finalMessage,
+      }),
+    );
+
+    const stream = await invokeWrappedStream(baseFn);
+    await stream.result();
+
+    expect(finalToolCallA.name).toBe("read");
+    expect(finalToolCallB.name).toBe("write");
+    expect(finalToolCallA.id).toBe("edit:22");
+    expect(finalToolCallB.id).toBe("call_auto_1");
+  });
 });
 
 describe("wrapStreamFnRepairMalformedToolCallArguments", () => {

src/agents/pi-embedded-runner/run/attempt.ts

@@ -667,6 +667,7 @@ function normalizeToolCallIdsInMessage(message: unknown): void {
   }
 
   let fallbackIndex = 1;
+  const assignedIds = new Set<string>();
   for (const block of content) {
     if (!block || typeof block !== "object") {
       continue;
@@ -678,20 +679,23 @@ function normalizeToolCallIdsInMessage(message: unknown): void {
     if (typeof typedBlock.id === "string") {
       const trimmedId = typedBlock.id.trim();
       if (trimmedId) {
-        if (typedBlock.id !== trimmedId) {
+        if (!assignedIds.has(trimmedId)) {
-          typedBlock.id = trimmedId;
+          if (typedBlock.id !== trimmedId) {
+            typedBlock.id = trimmedId;
+          }
+          assignedIds.add(trimmedId);
+          continue;
         }
-        usedIds.add(trimmedId);
-        continue;
       }
     }
 
     let fallbackId = "";
-    while (!fallbackId || usedIds.has(fallbackId)) {
+    while (!fallbackId || usedIds.has(fallbackId) || assignedIds.has(fallbackId)) {
       fallbackId = `call_auto_${fallbackIndex++}`;
     }
     typedBlock.id = fallbackId;
     usedIds.add(fallbackId);
+    assignedIds.add(fallbackId);
   }
 }
 

src/agents/tool-call-id.test.ts

@@ -29,6 +29,54 @@ const buildDuplicateIdCollisionInput = () =>
     },
   ]);
 
+const buildRepeatedRawIdInput = () =>
+  castAgentMessages([
+    {
+      role: "assistant",
+      content: [
+        { type: "toolCall", id: "edit:22", name: "edit", arguments: {} },
+        { type: "toolCall", id: "edit:22", name: "edit", arguments: {} },
+      ],
+    },
+    {
+      role: "toolResult",
+      toolCallId: "edit:22",
+      toolName: "edit",
+      content: [{ type: "text", text: "one" }],
+    },
+    {
+      role: "toolResult",
+      toolCallId: "edit:22",
+      toolName: "edit",
+      content: [{ type: "text", text: "two" }],
+    },
+  ]);
+
+const buildRepeatedSharedToolResultIdInput = () =>
+  castAgentMessages([
+    {
+      role: "assistant",
+      content: [
+        { type: "toolCall", id: "edit:22", name: "edit", arguments: {} },
+        { type: "toolCall", id: "edit:22", name: "edit", arguments: {} },
+      ],
+    },
+    {
+      role: "toolResult",
+      toolCallId: "edit:22",
+      toolUseId: "edit:22",
+      toolName: "edit",
+      content: [{ type: "text", text: "one" }],
+    },
+    {
+      role: "toolResult",
+      toolCallId: "edit:22",
+      toolUseId: "edit:22",
+      toolName: "edit",
+      content: [{ type: "text", text: "two" }],
+    },
+  ]);
+
 function expectCollisionIdsRemainDistinct(
   out: AgentMessage[],
   mode: "strict" | "strict9",
@@ -111,6 +159,26 @@ describe("sanitizeToolCallIdsForCloudCodeAssist", () => {
       expectCollisionIdsRemainDistinct(out, "strict");
     });
 
+    it("reuses one rewritten id when a tool result carries matching toolCallId and toolUseId", () => {
+      const input = buildRepeatedSharedToolResultIdInput();
+
+      const out = sanitizeToolCallIdsForCloudCodeAssist(input);
+      expect(out).not.toBe(input);
+      const { aId, bId } = expectCollisionIdsRemainDistinct(out, "strict");
+      const r1 = out[1] as Extract<AgentMessage, { role: "toolResult" }> & { toolUseId?: string };
+      const r2 = out[2] as Extract<AgentMessage, { role: "toolResult" }> & { toolUseId?: string };
+      expect(r1.toolUseId).toBe(aId);
+      expect(r2.toolUseId).toBe(bId);
+    });
+
+    it("assigns distinct IDs when identical raw tool call ids repeat", () => {
+      const input = buildRepeatedRawIdInput();
+
+      const out = sanitizeToolCallIdsForCloudCodeAssist(input);
+      expect(out).not.toBe(input);
+      expectCollisionIdsRemainDistinct(out, "strict");
+    });
+
     it("caps tool call IDs at 40 chars while preserving uniqueness", () => {
       const longA = `call_${"a".repeat(60)}`;
       const longB = `call_${"a".repeat(59)}b`;
@@ -181,6 +249,16 @@ describe("sanitizeToolCallIdsForCloudCodeAssist", () => {
       expect(aId).not.toMatch(/[_-]/);
       expect(bId).not.toMatch(/[_-]/);
     });
+
+    it("assigns distinct strict IDs when identical raw tool call ids repeat", () => {
+      const input = buildRepeatedRawIdInput();
+
+      const out = sanitizeToolCallIdsForCloudCodeAssist(input, "strict");
+      expect(out).not.toBe(input);
+      const { aId, bId } = expectCollisionIdsRemainDistinct(out, "strict");
+      expect(aId).not.toMatch(/[_-]/);
+      expect(bId).not.toMatch(/[_-]/);
+    });
   });
 
   describe("strict9 mode (Mistral tool call IDs)", () => {
@@ -231,5 +309,27 @@ describe("sanitizeToolCallIdsForCloudCodeAssist", () => {
       expect(aId.length).toBe(9);
       expect(bId.length).toBe(9);
     });
+
+    it("assigns distinct strict9 IDs when identical raw tool call ids repeat", () => {
+      const input = buildRepeatedRawIdInput();
+
+      const out = sanitizeToolCallIdsForCloudCodeAssist(input, "strict9");
+      expect(out).not.toBe(input);
+      const { aId, bId } = expectCollisionIdsRemainDistinct(out, "strict9");
+      expect(aId.length).toBe(9);
+      expect(bId.length).toBe(9);
+    });
+
+    it("reuses one rewritten strict9 id when a tool result carries matching toolCallId and toolUseId", () => {
+      const input = buildRepeatedSharedToolResultIdInput();
+
+      const out = sanitizeToolCallIdsForCloudCodeAssist(input, "strict9");
+      expect(out).not.toBe(input);
+      const { aId, bId } = expectCollisionIdsRemainDistinct(out, "strict9");
+      const r1 = out[1] as Extract<AgentMessage, { role: "toolResult" }> & { toolUseId?: string };
+      const r2 = out[2] as Extract<AgentMessage, { role: "toolResult" }> & { toolUseId?: string };
+      expect(r1.toolUseId).toBe(aId);
+      expect(r2.toolUseId).toBe(bId);
+    });
   });
 });

src/agents/tool-call-id.ts

@@ -144,9 +144,55 @@ function makeUniqueToolId(params: { id: string; used: Set<string>; mode: ToolCal
   return `${candidate.slice(0, MAX_LEN - ts.length)}${ts}`;
 }
 
+function createOccurrenceAwareResolver(mode: ToolCallIdMode): {
+  resolveAssistantId: (id: string) => string;
+  resolveToolResultId: (id: string) => string;
+} {
+  const used = new Set<string>();
+  const assistantOccurrences = new Map<string, number>();
+  const orphanToolResultOccurrences = new Map<string, number>();
+  const pendingByRawId = new Map<string, string[]>();
+
+  const allocate = (seed: string): string => {
+    const next = makeUniqueToolId({ id: seed, used, mode });
+    used.add(next);
+    return next;
+  };
+
+  const resolveAssistantId = (id: string): string => {
+    const occurrence = (assistantOccurrences.get(id) ?? 0) + 1;
+    assistantOccurrences.set(id, occurrence);
+    const next = allocate(occurrence === 1 ? id : `${id}:${occurrence}`);
+    const pending = pendingByRawId.get(id);
+    if (pending) {
+      pending.push(next);
+    } else {
+      pendingByRawId.set(id, [next]);
+    }
+    return next;
+  };
+
+  const resolveToolResultId = (id: string): string => {
+    const pending = pendingByRawId.get(id);
+    if (pending && pending.length > 0) {
+      const next = pending.shift()!;
+      if (pending.length === 0) {
+        pendingByRawId.delete(id);
+      }
+      return next;
+    }
+
+    const occurrence = (orphanToolResultOccurrences.get(id) ?? 0) + 1;
+    orphanToolResultOccurrences.set(id, occurrence);
+    return allocate(`${id}:tool_result:${occurrence}`);
+  };
+
+  return { resolveAssistantId, resolveToolResultId };
+}
+
 function rewriteAssistantToolCallIds(params: {
   message: Extract<AgentMessage, { role: "assistant" }>;
-  resolve: (id: string) => string;
+  resolveId: (id: string) => string;
 }): Extract<AgentMessage, { role: "assistant" }> {
   const content = params.message.content;
   if (!Array.isArray(content)) {
@@ -168,7 +214,7 @@ function rewriteAssistantToolCallIds(params: {
     ) {
       return block;
     }
-    const nextId = params.resolve(id);
+    const nextId = params.resolveId(id);
     if (nextId === id) {
       return block;
     }
@@ -184,7 +230,7 @@ function rewriteAssistantToolCallIds(params: {
 
 function rewriteToolResultIds(params: {
   message: Extract<AgentMessage, { role: "toolResult" }>;
-  resolve: (id: string) => string;
+  resolveId: (id: string) => string;
 }): Extract<AgentMessage, { role: "toolResult" }> {
   const toolCallId =
     typeof params.message.toolCallId === "string" && params.message.toolCallId
@@ -192,9 +238,14 @@ function rewriteToolResultIds(params: {
       : undefined;
   const toolUseId = (params.message as { toolUseId?: unknown }).toolUseId;
   const toolUseIdStr = typeof toolUseId === "string" && toolUseId ? toolUseId : undefined;
+  const sharedRawId =
+    toolCallId && toolUseIdStr && toolCallId === toolUseIdStr ? toolCallId : undefined;
 
-  const nextToolCallId = toolCallId ? params.resolve(toolCallId) : undefined;
+  const sharedResolvedId = sharedRawId ? params.resolveId(sharedRawId) : undefined;
-  const nextToolUseId = toolUseIdStr ? params.resolve(toolUseIdStr) : undefined;
+  const nextToolCallId =
+    sharedResolvedId ?? (toolCallId ? params.resolveId(toolCallId) : undefined);
+  const nextToolUseId =
+    sharedResolvedId ?? (toolUseIdStr ? params.resolveId(toolUseIdStr) : undefined);
 
   if (nextToolCallId === toolCallId && nextToolUseId === toolUseIdStr) {
     return params.message;
@@ -219,21 +270,11 @@ export function sanitizeToolCallIdsForCloudCodeAssist(
 ): AgentMessage[] {
   // Strict mode: only [a-zA-Z0-9]
   // Strict9 mode: only [a-zA-Z0-9], length 9 (Mistral tool call requirement)
-  // Sanitization can introduce collisions (e.g. `a|b` and `a:b` -> `ab`).
+  // Sanitization can introduce collisions, and some providers also reject raw
-  // Fix by applying a stable, transcript-wide mapping and de-duping via suffix.
+  // duplicate tool-call IDs. Track assistant occurrences in-order so repeated
-  const map = new Map<string, string>();
+  // raw IDs receive distinct rewritten IDs, while matching tool results consume
-  const used = new Set<string>();
+  // the same rewritten IDs in encounter order.
-
+  const { resolveAssistantId, resolveToolResultId } = createOccurrenceAwareResolver(mode);
-  const resolve = (id: string) => {
-    const existing = map.get(id);
-    if (existing) {
-      return existing;
-    }
-    const next = makeUniqueToolId({ id, used, mode });
-    map.set(id, next);
-    used.add(next);
-    return next;
-  };
 
   let changed = false;
   const out = messages.map((msg) => {
@@ -244,7 +285,7 @@ export function sanitizeToolCallIdsForCloudCodeAssist(
     if (role === "assistant") {
       const next = rewriteAssistantToolCallIds({
         message: msg as Extract<AgentMessage, { role: "assistant" }>,
-        resolve,
+        resolveId: resolveAssistantId,
       });
       if (next !== msg) {
         changed = true;
@@ -254,7 +295,7 @@ export function sanitizeToolCallIdsForCloudCodeAssist(
     if (role === "toolResult") {
       const next = rewriteToolResultIds({
         message: msg as Extract<AgentMessage, { role: "toolResult" }>,
-        resolve,
+        resolveId: resolveToolResultId,
       });
       if (next !== msg) {
         changed = true;

src/agents/transcript-policy.ts

@@ -78,7 +78,10 @@ export function resolveTranscriptPolicy(params: {
       provider,
       modelId,
     });
-  const requiresOpenAiCompatibleToolIdSanitization = params.modelApi === "openai-completions";
+  const requiresOpenAiCompatibleToolIdSanitization =
+    params.modelApi === "openai-completions" ||
+    (!isOpenAi &&
+      (params.modelApi === "openai-responses" || params.modelApi === "openai-codex-responses"));
 
   // Anthropic Claude endpoints can reject replayed `thinking` blocks unless the
   // original signatures are preserved byte-for-byte. Drop them at send-time to