fix(codex): strip orphaned tool outputs (#9228)

Validated in local merge-train (devbox-vm-06-dev002) @ combined-tip (FAST gates — only pre-existing audit.test.ts flake).

changelog.d/fixes/9228-codex-orphaned-tool-outputs.md

@@ -0,0 +1 @@
+- **fix(codex):** strip orphaned tool outputs from compacted conversations. (thanks @raflyazf)

open-sse/executors/codex.ts

@@ -307,6 +307,59 @@ export function stripStoredItemReferences(body: Record<string, unknown>): void {
   }
 }
 
+function stripOrphanedCodexFunctionCallOutputs(body: Record<string, unknown>): void {
+  if (!Array.isArray(body.input)) return;
+  // A previous_response_id delegates history resolution to the upstream
+  // Responses service, so a matching function_call may legitimately live in
+  // that remote response rather than in the local input array.
+  if (typeof body.previous_response_id === "string" && body.previous_response_id.trim()) return;
+
+  const callIds = new Set<string>();
+  let outputCount = 0;
+
+  for (const item of body.input) {
+    if (!item || typeof item !== "object" || Array.isArray(item)) continue;
+    const record = item as Record<string, unknown>;
+
+    if (record.type === "function_call" && typeof record.call_id === "string") {
+      callIds.add(record.call_id);
+    }
+
+    if (Array.isArray(record.tool_calls)) {
+      for (const toolCall of record.tool_calls) {
+        if (!toolCall || typeof toolCall !== "object" || Array.isArray(toolCall)) continue;
+        const toolCallId = (toolCall as Record<string, unknown>).id;
+        if (typeof toolCallId === "string") {
+          callIds.add(toolCallId);
+        }
+      }
+    }
+
+    if (record.type === "function_call_output") {
+      outputCount++;
+    }
+  }
+
+  if (outputCount === 0) return;
+
+  const before = body.input.length;
+  body.input = body.input.filter((item) => {
+    if (!item || typeof item !== "object" || Array.isArray(item)) return true;
+    const record = item as Record<string, unknown>;
+    if (record.type === "function_call_output" && typeof record.call_id === "string") {
+      return callIds.has(record.call_id);
+    }
+    return true;
+  });
+
+  const removedCount = before - body.input.length;
+  if (removedCount > 0) {
+    console.debug(
+      `[Codex] stripOrphanedCodexFunctionCallOutputs: removed ${removedCount} orphaned function_call_output item(s)`
+    );
+  }
+}
+
 function repairMissingCodexFunctionCallOutputs(body: Record<string, unknown>): void {
   if (!Array.isArray(body.input)) return;
 
@@ -1293,6 +1346,7 @@ export class CodexExecutor extends BaseExecutor {
         dropInternalAssistantMessages: !nativeCodexPassthrough,
       });
     }
+    stripOrphanedCodexFunctionCallOutputs(body);
     repairMissingCodexFunctionCallOutputs(body);
 
     // ── Cache-aware system prompt handling (both paths) ──

tests/unit/codex-orphaned-tool-outputs-2928.test.ts

@@ -0,0 +1,137 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+import { mkdtempSync, rmSync } from "node:fs";
+import { tmpdir } from "node:os";
+import { join } from "node:path";
+
+const TEST_DATA_DIR = mkdtempSync(join(tmpdir(), "omniroute-codex-orphaned-2928-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const core = await import("../../src/lib/db/core.ts");
+const { CodexExecutor } = await import("../../open-sse/executors/codex.ts");
+
+type InputItem = Record<string, unknown>;
+
+function transform(input: InputItem[]): InputItem[] {
+  const executor = new CodexExecutor();
+  const result = executor.transformRequest(
+    "gpt-5.6-sol",
+    {
+      _nativeCodexPassthrough: true,
+      model: "gpt-5.6-sol",
+      input,
+      stream: true,
+    },
+    true,
+    { requestEndpointPath: "/responses" }
+  );
+
+  assert.ok(Array.isArray(result.input));
+  return result.input as InputItem[];
+}
+
+function toolOutputs(input: InputItem[]): InputItem[] {
+  return input.filter((item) => item.type === "function_call_output");
+}
+
+test.after(() => {
+  core.resetDbInstance();
+  rmSync(TEST_DATA_DIR, { recursive: true, force: true });
+});
+
+test("Codex strips function_call_output items without matching function calls", () => {
+  const result = transform([
+    {
+      type: "message",
+      role: "user",
+      content: [{ type: "input_text", text: "hi" }],
+    },
+    { type: "function_call_output", call_id: "call_orphan1", output: "result1" },
+    { type: "function_call_output", call_id: "call_orphan2", output: "result2" },
+  ]);
+
+  assert.deepEqual(toolOutputs(result), []);
+});
+
+test("Codex keeps a function_call_output with a matching Responses function_call", () => {
+  const result = transform([
+    {
+      type: "function_call",
+      call_id: "call_valid",
+      name: "read_file",
+      arguments: '{"path":"a.txt"}',
+    },
+    { type: "function_call_output", call_id: "call_valid", output: "file contents" },
+  ]);
+
+  assert.deepEqual(toolOutputs(result), [
+    { type: "function_call_output", call_id: "call_valid", output: "file contents" },
+  ]);
+});
+
+test("Codex keeps matched outputs and removes orphaned outputs from mixed input", () => {
+  const result = transform([
+    { type: "function_call", call_id: "call_keep", name: "read_file", arguments: "{}" },
+    { type: "function_call_output", call_id: "call_keep", output: "ok" },
+    { type: "function_call_output", call_id: "call_orphan", output: "orphaned" },
+  ]);
+
+  assert.deepEqual(toolOutputs(result), [
+    { type: "function_call_output", call_id: "call_keep", output: "ok" },
+  ]);
+});
+
+test("Codex recognizes embedded Chat Completions tool_calls as matching calls", () => {
+  const result = transform([
+    {
+      type: "message",
+      role: "assistant",
+      content: [],
+      tool_calls: [
+        { id: "call_cc1", type: "function", function: { name: "read_file", arguments: "{}" } },
+      ],
+    },
+    { type: "function_call_output", call_id: "call_cc1", output: "ok" },
+    { type: "function_call_output", call_id: "call_cc2", output: "orphaned" },
+  ]);
+
+  assert.deepEqual(toolOutputs(result), [
+    { type: "function_call_output", call_id: "call_cc1", output: "ok" },
+  ]);
+});
+
+test("Codex preserves tool outputs linked to a remote previous_response_id", () => {
+  const executor = new CodexExecutor();
+  const result = executor.transformRequest(
+    "gpt-5.6-sol",
+    {
+      _nativeCodexPassthrough: true,
+      previous_response_id: "resp_remote_history",
+      input: [{ type: "function_call_output", call_id: "call_remote", output: "remote result" }],
+      stream: true,
+    },
+    true,
+    { requestEndpointPath: "/responses" }
+  );
+
+  assert.deepEqual(result.input, [
+    { type: "function_call_output", call_id: "call_remote", output: "remote result" },
+  ]);
+});
+
+test("Codex leaves inputs without function_call_output items unchanged", () => {
+  const input = [
+    {
+      type: "message",
+      role: "user",
+      content: [{ type: "input_text", text: "hi" }],
+    },
+    { type: "function_call", call_id: "call_1", name: "read_file", arguments: "{}" },
+  ];
+
+  assert.deepEqual(transform(input), [
+    input[0],
+    input[1],
+    { type: "function_call_output", call_id: "call_1", output: "" },
+  ]);
+});