fix: add reverse normalization max_tokens/max_completion_tokens → max_output_tokens for Responses targets

Per review feedback: when targeting openai-responses, also normalize
max_tokens and max_completion_tokens INTO max_output_tokens, not just
skip the outbound normalization. This covers the case where a Chat
Completions client sends max_tokens to a Responses endpoint via
passthrough.

Extends test to cover both reverse normalization paths.

open-sse/handlers/chatCore.ts

@@ -910,12 +910,21 @@ export async function handleChatCore({
   }
 
   // ── Common input sanitization (runs for ALL paths including passthrough) ──
-  // #994: Normalize max_output_tokens to max_tokens for universal compatibility.
-  // Skip for Responses API targets where max_output_tokens is the canonical field;
-  // normalizing it to max_tokens there causes "Unsupported parameter: max_tokens"
-  // errors because Responses→Responses passthrough skips the translator that would
-  // convert it back (see openaiToOpenAIResponsesRequest).
-  if (body.max_output_tokens !== undefined && targetFormat !== FORMATS.OPENAI_RESPONSES) {
+  // #994: Normalize between max_output_tokens and max_tokens for universal compatibility.
+  // For Responses API targets, max_output_tokens is the canonical field. For others,
+  // max_tokens is preferred. We handle normalization here to support passthrough
+  // paths where the translator is skipped.
+  if (targetFormat === FORMATS.OPENAI_RESPONSES) {
+    if (body.max_output_tokens === undefined) {
+      if (body.max_completion_tokens !== undefined) {
+        body.max_output_tokens = body.max_completion_tokens;
+        delete body.max_completion_tokens;
+      } else if (body.max_tokens !== undefined) {
+        body.max_output_tokens = body.max_tokens;
+        delete body.max_tokens;
+      }
+    }
+  } else if (body.max_output_tokens !== undefined) {
     if (body.max_tokens === undefined) {
       body.max_tokens = body.max_output_tokens;
     }

tests/unit/chatcore-sanitization.test.ts

@@ -209,6 +209,58 @@ test("chatCore sanitization preserves max_output_tokens for openai-responses tar
 
   // max_output_tokens should survive sanitization for Responses targets
   assert.equal("max_tokens" in call.body, false, "max_tokens must not be injected for Responses targets");
+
+  // Reverse normalization: max_tokens → max_output_tokens for Responses targets
+  const fromMaxTokens = await invokeChatCore({
+    endpoint: "/v1/responses",
+    body: {
+      model: "gpt-5.4",
+      max_tokens: 2048,
+      input: [{ role: "user", content: "hello" }],
+    },
+    responseFactory: () =>
+      new Response(
+        JSON.stringify({
+          id: "resp_test2",
+          object: "response",
+          status: "completed",
+          model: "gpt-5.4",
+          output: [
+            { type: "message", role: "assistant", content: [{ type: "output_text", text: "ok" }] },
+          ],
+          usage: { input_tokens: 1, output_tokens: 1, total_tokens: 2 },
+        }),
+        { status: 200, headers: { "Content-Type": "application/json" } }
+      ),
+  });
+
+  assert.equal("max_tokens" in fromMaxTokens.call.body, false, "max_tokens should be converted to max_output_tokens");
+
+  // Reverse normalization: max_completion_tokens → max_output_tokens for Responses targets
+  const fromMaxCompletion = await invokeChatCore({
+    endpoint: "/v1/responses",
+    body: {
+      model: "gpt-5.4",
+      max_completion_tokens: 8192,
+      input: [{ role: "user", content: "hello" }],
+    },
+    responseFactory: () =>
+      new Response(
+        JSON.stringify({
+          id: "resp_test3",
+          object: "response",
+          status: "completed",
+          model: "gpt-5.4",
+          output: [
+            { type: "message", role: "assistant", content: [{ type: "output_text", text: "ok" }] },
+          ],
+          usage: { input_tokens: 1, output_tokens: 1, total_tokens: 2 },
+        }),
+        { status: 200, headers: { "Content-Type": "application/json" } }
+      ),
+  });
+
+  assert.equal("max_completion_tokens" in fromMaxCompletion.call.body, false, "max_completion_tokens should be converted to max_output_tokens");
 });
 
 test("chatCore sanitization strips empty message names and filters empty tool names", async () => {