fix(responses): never send prompt_cache_retention to gpt-5.6 models

package.json

@@ -1,6 +1,6 @@
 {
   "name": "@bitkyc08/opencodex",
-  "version": "2.24.2",
+  "version": "2.25.0",
   "description": "Universal provider proxy for OpenAI Codex & Claude Code — use any LLM with Codex CLI/App/SDK and Claude Code",
   "type": "module",
   "main": "./bin/package-main.mjs",

src/adapters/openai-responses.ts

@@ -815,6 +815,28 @@ function stripUnsupportedForwardParams(body: unknown): unknown {
   return rest;
 }
 
+/**
+ * The ChatGPT codex backend rejects `prompt_cache_retention` with
+ * `{"detail":"Unsupported parameter: prompt_cache_retention"}` on gpt-5.6
+ * models (gpt-5.6-luna / gpt-5.6-sol), aborting the whole agent turn mid-run.
+ * The parameter is emitted client-side by some Codex App builds - it does not
+ * exist anywhere in codex-rs - and request bodies are forwarded opaquely, so
+ * one bad field kills the turn (#2092).
+ *
+ * Other models keep the field: the backend's cache handling is account-level
+ * and has provably varied by deployment (one accepted "24h" and echoed it
+ * back), so stripping globally would silently drop a parameter a deployment
+ * honors. Model-scoped, matching the report's invariant: never send it to
+ * gpt-5.6.
+ */
+function stripPromptCacheRetentionForGpt56(body: unknown, modelId: string | undefined): unknown {
+  if (modelId === undefined || !modelId.startsWith("gpt-5.6")) return body;
+  if (!isPlainObject(body)) return body;
+  if (!Object.prototype.hasOwnProperty.call(body, "prompt_cache_retention")) return body;
+  const { prompt_cache_retention: _pcr, ...rest } = body;
+  return rest;
+}
+
 const IMAGE_GEN_NAMESPACE = "image_gen";
 const HOSTED_IMAGE_GENERATION_TOOL = "image_generation";
 const IMAGE_GEN_DOTTED_PREFIX = `${IMAGE_GEN_NAMESPACE}.`;
@@ -1382,6 +1404,7 @@ export function createResponsesPassthroughAdapter(provider: OcxProviderConfig):
       }
       if (forward) {
         outBody = stripUnsupportedForwardParams(outBody);
+        outBody = stripPromptCacheRetentionForGpt56(outBody, parsed.modelId);
       } else {
         outBody = preferConfiguredHostedTools(
           outBody,

tests/openai-responses-passthrough.test.ts

@@ -1982,6 +1982,49 @@ describe("OpenAI Responses forward-mode unsupported param stripping", () => {
     expect(body.model).toBe("gpt-5.6-sol");
   });
 
+
+
+  test("forward mode strips prompt_cache_retention for gpt-5.6 models (#2092)", () => {
+    const adapter = createResponsesPassthroughAdapter(provider);
+    for (const modelId of ["gpt-5.6-luna", "gpt-5.6-sol"]) {
+      const request = adapter.buildRequest({
+        modelId,
+        context: { messages: [] },
+        stream: true,
+        options: {},
+        _rawBody: {
+          model: modelId,
+          input: [{ role: "user", content: [{ type: "input_text", text: "ping" }] }],
+          stream: true,
+          store: false,
+          prompt_cache_retention: "24h",
+        },
+      }, meta);
+      const body = JSON.parse(request.body) as Record<string, unknown>;
+      expect(body).not.toHaveProperty("prompt_cache_retention");
+      expect(body.model).toBe(modelId);
+    }
+  });
+
+  test("forward mode keeps prompt_cache_retention for non-gpt-5.6 models (#2092)", () => {
+    const adapter = createResponsesPassthroughAdapter(provider);
+    const request = adapter.buildRequest({
+      modelId: "gpt-5.5",
+      context: { messages: [] },
+      stream: true,
+      options: {},
+      _rawBody: {
+        model: "gpt-5.5",
+        input: "hi",
+        prompt_cache_retention: "24h",
+      },
+    }, meta);
+    const body = JSON.parse(request.body) as Record<string, unknown>;
+    // The backend's cache handling varies by deployment and has accepted this
+    // field before; only gpt-5.6 provably rejects it, so it stays.
+    expect(body.prompt_cache_retention).toBe("24h");
+  });
+
   test("forward mode is a no-op when neither field is present", () => {
     const adapter = createResponsesPassthroughAdapter(provider);
     const { max_output_tokens: _m, metadata: _d, ...codexBody } = rawBody;