fix(sse): echo the client-requested model, not the combo leaf (#18494)

changelog.d/fixes/18495-echo-client-requested-model.md

@@ -0,0 +1 @@
+- **fix(sse):** the response-model echo (`echoRequestedModelName`, Codex CLI, Claude Code) now returns the model the client requested instead of the combo leaf target ([#18495](https://github.com/cheaper-inference/OmniRoute/pull/18495), closes [#18494](https://github.com/cheaper-inference/OmniRoute/issues/18494)) — thanks @xiaoyaner0201

open-sse/handlers/chatCore.ts

@@ -4,7 +4,7 @@ import {
   injectMemoryAndSkills,
   mergeInjectedFallbackOwnerNames,
 } from "./chatCore/memorySkillsInjection.ts";
-import { resolveChatCoreRequestSetup } from "./chatCore/requestSetup.ts";
+import { resolveChatCoreRequestSetup, resolveClientEchoModel } from "./chatCore/requestSetup.ts";
 import { shouldSkipNativeCodexReactiveCompaction } from "./chatCore/nativeCodexCompaction.ts";
 import {
   normalizeOpenAICompatibleTools,
@@ -867,11 +867,12 @@ async function handleChatCoreInner({
   // `opencode-zen/nemotron-3-ultra-free`) is not recognized by the client on `--resume`.
   const isClaudeCodeClient = isClaudeCodeOriginatedHeaders(clientRawRequest?.headers);
 
+  const clientEchoModel = resolveClientEchoModel(clientRawRequest?.body, requestedModel);
   let echoModel =
     (settings.echoRequestedModelName === true || isCodexResponsesEcho || isClaudeCodeClient) &&
-    typeof requestedModel === "string" &&
-    requestedModel
-      ? requestedModel
+    typeof clientEchoModel === "string" &&
+    clientEchoModel
+      ? clientEchoModel
       : null;
   // Auto-echo the listing-valid form for bare requests to noAuth catalog
   // providers so clients validating response.model against /v1/models don't warn.

open-sse/handlers/chatCore/requestSetup.ts

@@ -49,3 +49,19 @@ export function resolveChatCoreRequestSetup(
     typeof body?.model === "string" && body.model.trim().length > 0 ? body.model : model;
   return { apiFormat, customModelTargetFormat, requestedModel };
 }
+
+/**
+ * The model name the response-model echo (#1311 opt-in, #3697 Codex CLI, Claude Code) writes
+ * back. Combo and alias dispatch rewrite the dispatched body.model to the resolved leaf target
+ * before handleChatCore runs, so requestedModel is the leaf there. clientRawRequest.body is the
+ * snapshot of what the client actually sent; prefer its model and fall back to requestedModel.
+ */
+export function resolveClientEchoModel(clientRawBody: unknown, requestedModel: string): string {
+  const clientModel =
+    clientRawBody && typeof clientRawBody === "object"
+      ? (clientRawBody as { model?: unknown }).model
+      : undefined;
+  return typeof clientModel === "string" && clientModel.trim().length > 0
+    ? clientModel
+    : requestedModel;
+}

tests/unit/response-model-echo-client-requested.test.ts

@@ -0,0 +1,100 @@
+// The response-model echo (#1311 opt-in, #3697 Codex CLI) must echo the model name the
+// client sent. Combo and alias dispatch rewrite the dispatched body.model to the resolved
+// leaf target (chatHelpers.ts sends `${provider}/${model}`) before handleChatCore runs, so
+// echoing body.model leaks the leaf target instead of the alias/combo the client asked for.
+// clientRawRequest.body is the snapshot of what the client actually sent.
+import test from "node:test";
+import assert from "node:assert/strict";
+import fs from "node:fs";
+import os from "node:os";
+import path from "node:path";
+
+const TEST_DATA_DIR = fs.mkdtempSync(path.join(os.tmpdir(), "omniroute-echo-client-model-"));
+process.env.DATA_DIR = TEST_DATA_DIR;
+
+const { handleChatCore } = await import("../../open-sse/handlers/chatCore.ts");
+const settingsDb = await import("../../src/lib/db/settings.ts");
+const core = await import("../../src/lib/db/core.ts");
+
+const CLIENT_MODEL = "team/my-alias-combo";
+const LEAF_MODEL = "gpt-4o-mini";
+
+function noopLog() {
+  return { debug() {}, info() {}, warn() {}, error() {} };
+}
+
+async function invoke({
+  stream,
+  echoSetting = true,
+  userAgent = "unit-test",
+  clientModel = CLIENT_MODEL,
+}) {
+  await settingsDb.updateSettings({ echoRequestedModelName: echoSetting });
+  const originalFetch = globalThis.fetch;
+  globalThis.fetch = async () =>
+    stream
+      ? new Response(
+          `data: {"id":"c1","object":"chat.completion.chunk","model":"${LEAF_MODEL}","choices":[{"index":0,"delta":{"role":"assistant","content":"ok"}}]}\n\n` +
+            `data: {"id":"c1","object":"chat.completion.chunk","model":"${LEAF_MODEL}","choices":[{"index":0,"delta":{},"finish_reason":"stop"}]}\n\ndata: [DONE]\n\n`,
+          { status: 200, headers: { "Content-Type": "text/event-stream" } }
+        )
+      : new Response(
+          JSON.stringify({
+            id: "c1",
+            object: "chat.completion",
+            model: LEAF_MODEL,
+            choices: [
+              { index: 0, message: { role: "assistant", content: "ok" }, finish_reason: "stop" },
+            ],
+            usage: { prompt_tokens: 1, completion_tokens: 1, total_tokens: 2 },
+          }),
+          { status: 200, headers: { "Content-Type": "application/json" } }
+        );
+  try {
+    const messages = [{ role: "user", content: "hello" }];
+    const result = await handleChatCore({
+      // What combo/alias dispatch hands to chatCore: body.model already rewritten to the leaf.
+      body: { model: `openai/${LEAF_MODEL}`, stream, messages },
+      modelInfo: { provider: "openai", model: LEAF_MODEL, extendedContext: false },
+      credentials: { apiKey: "sk-test", providerSpecificData: {} },
+      log: noopLog(),
+      clientRawRequest: {
+        endpoint: "/v1/chat/completions",
+        body: { model: clientModel, stream, messages },
+        headers: new Headers({
+          accept: stream ? "text/event-stream" : "application/json",
+          "user-agent": userAgent,
+        }),
+      },
+      userAgent,
+    });
+    assert.equal(result.success, true, JSON.stringify(result).slice(0, 300));
+    const text = await result.response.text();
+    return [...text.matchAll(/"model":"([^"]+)"/g)].map((m) => m[1]);
+  } finally {
+    globalThis.fetch = originalFetch;
+  }
+}
+
+test.after(() => {
+  try {
+    core.getDbInstance().close();
+  } catch {}
+  fs.rmSync(TEST_DATA_DIR, { recursive: true, force: true, maxRetries: 5, retryDelay: 100 });
+});
+
+test("echo returns the client-requested model, not the rewritten leaf (non-streaming)", async () => {
+  const models = await invoke({ stream: false });
+  assert.deepEqual([...new Set(models)], [CLIENT_MODEL]);
+});
+
+test("echo returns the client-requested model, not the rewritten leaf (streaming)", async () => {
+  const models = await invoke({ stream: true });
+  assert.ok(models.length > 0);
+  assert.deepEqual([...new Set(models)], [CLIENT_MODEL]);
+});
+
+test("echo off: response model is not rewritten to the client name", async () => {
+  const models = await invoke({ stream: false, echoSetting: false });
+  assert.ok(!models.includes(CLIENT_MODEL), JSON.stringify(models));
+});