fix(sse): stop Codex/Responses sanitizer turning system image_url into output_text (#8089) (#8147)

changelog.d/fixes/8089-responses-system-image.md

@@ -0,0 +1 @@
+- fix(sse): stop Codex/Responses sanitizer from turning system/developer image_url parts into invalid output_text (#8089)

open-sse/services/responsesInputSanitizer.ts

@@ -67,7 +67,11 @@ function sanitizeContentPart(part: unknown, role: string): unknown {
 
   if (record.type === "image_url") {
     const url = imageUrlToText(record.image_url);
-    if (role === "user") {
+    // `output_text` is only a legal content-part type on assistant-role OUTPUT
+    // items. Every other role (user, system, developer, ...) is input-side and
+    // must use `input_image` -- otherwise the Codex/Responses backend rejects
+    // the replayed history with "Invalid value: 'output_text'" (#8089).
+    if (role !== "assistant") {
       const next: JsonRecord = { type: "input_image", image_url: url };
       const image = toRecord(record.image_url);
       if (image?.detail !== undefined) next.detail = image.detail;

tests/unit/codex-system-message-image-8089.test.ts

@@ -0,0 +1,89 @@
+import test from "node:test";
+import assert from "node:assert/strict";
+
+import { sanitizeResponsesInputItems } from "../../open-sse/services/responsesInputSanitizer.ts";
+import { CodexExecutor } from "../../open-sse/executors/codex.ts";
+
+test("[repro #8089] mid-conversation system-role message with image_url must NOT become output_text", () => {
+  const input = [
+    {
+      type: "message",
+      role: "system",
+      content: [
+        { type: "text", text: "Reminder: verify the diff before committing." },
+        { type: "image_url", image_url: { url: "https://example.com/diff-screenshot.png" } },
+      ],
+    },
+  ];
+
+  const sanitized = sanitizeResponsesInputItems(input, true, {}) as Array<{
+    content: Array<{ type: string }>;
+  }>;
+  const imagePart = sanitized[0].content.find(
+    (p) => p.type === "output_text" || p.type === "input_image"
+  );
+
+  assert.ok(imagePart);
+  assert.equal(imagePart?.type, "input_image");
+});
+
+test("[repro #8089] developer-role message with image_url must NOT become output_text", () => {
+  const input = [
+    {
+      type: "message",
+      role: "developer",
+      content: [{ type: "image_url", image_url: { url: "https://example.com/diff-screenshot.png" } }],
+    },
+  ];
+
+  const sanitized = sanitizeResponsesInputItems(input, true, {}) as Array<{
+    content: Array<{ type: string }>;
+  }>;
+
+  assert.equal(sanitized[0].content[0].type, "input_image");
+});
+
+test("assistant-role message with image_url still becomes output_text (unchanged behavior)", () => {
+  const input = [
+    {
+      type: "message",
+      role: "assistant",
+      content: [{ type: "image_url", image_url: { url: "https://example.com/diff-screenshot.png" } }],
+    },
+  ];
+
+  const sanitized = sanitizeResponsesInputItems(input, true, {}) as Array<{
+    content: Array<{ type: string }>;
+  }>;
+
+  assert.equal(sanitized[0].content[0].type, "output_text");
+});
+
+test("Codex native passthrough: mid-conversation system message with image content never serializes output_text (#8089)", () => {
+  const executor = new CodexExecutor();
+  const body = {
+    _nativeCodexPassthrough: true,
+    input: [
+      {
+        type: "message",
+        role: "system",
+        content: [
+          { type: "text", text: "Reminder: verify the diff before committing." },
+          { type: "image_url", image_url: { url: "https://example.com/diff-screenshot.png" } },
+        ],
+      },
+      {
+        type: "message",
+        role: "user",
+        content: [{ type: "input_text", text: "What do you see?" }],
+      },
+    ],
+    stream: false,
+  };
+
+  const result = executor.transformRequest("gpt-5.6", body, false, {
+    requestEndpointPath: "/responses",
+  });
+
+  assert.equal(JSON.stringify(result.input).includes('"type":"output_text"'), false);
+});