From dc2db9ca2123a2ad0ef44062b323923e232c3215 Mon Sep 17 00:00:00 2001
From: Alexsander Hamir <alexsanderhamirgomesbaptista@gmail.com>
Date: Mon, 2 Feb 2026 12:51:31 -0800
Subject: [PATCH] feat(proxy): start mock chat API once in main process on port
 8090

- Add ProxyInitializationHelpers._start_mock_chat_api_once() in proxy_cli
- Start mock (chat_completions_api.app) in a daemon thread on 8090 before forking workers
- Remove mock startup from proxy_server lifespan so workers never start it
- Single mock instance regardless of --num_workers; no port conflict
---
 litellm/proxy/proxy_cli.py    | 33 +++++++++++++++++++++++++++++++++
 litellm/proxy/proxy_server.py |  2 ++
 2 files changed, 35 insertions(+)

diff --git a/litellm/proxy/proxy_cli.py b/litellm/proxy/proxy_cli.py
index 2bc1c8f8e98..e70112af41b 100644
--- a/litellm/proxy/proxy_cli.py
+++ b/litellm/proxy/proxy_cli.py
@@ -5,6 +5,7 @@ import os
 import random
 import subprocess
 import sys
+import threading
 import urllib.parse as urlparse
 from typing import TYPE_CHECKING, Any, Optional, Union
 
@@ -116,6 +117,35 @@ class ProxyInitializationHelpers:
         )
         print(completion_response)  # noqa
 
+    @staticmethod
+    def _start_mock_chat_api_once() -> None:
+        """Start the mock chat completions API on port 8090 once in the main process (for testing)."""
+        _proxy_cli_dir = os.path.dirname(os.path.abspath(__file__))
+        _litellm_root = os.path.dirname(_proxy_cli_dir)
+        _repo_root = os.path.dirname(_litellm_root)
+        if _repo_root not in sys.path:
+            sys.path.insert(0, _repo_root)
+        try:
+            from chat_completions_api import app as mock_chat_app
+        except ImportError:
+            return
+
+        def _run_mock():
+            import uvicorn
+
+            uvicorn.run(
+                mock_chat_app,
+                host="0.0.0.0",
+                port=8090,
+                log_level="warning",
+            )
+
+        t = threading.Thread(target=_run_mock, daemon=True)
+        t.start()
+        print(  # noqa: T201
+            "\033[1;32mLiteLLM: Mock chat completions API started on http://0.0.0.0:8090 (for testing)\033[0m\n"
+        )
+
     @staticmethod
     def _get_default_unvicorn_init_args(
         host: str,
@@ -826,6 +856,9 @@ def run_server(  # noqa: PLR0915
             )
             return
 
+        # Start mock chat API once in main process (port 8090) before forking workers
+        ProxyInitializationHelpers._start_mock_chat_api_once()
+
         uvicorn_args = ProxyInitializationHelpers._get_default_unvicorn_init_args(
             host=host,
             port=port,
diff --git a/litellm/proxy/proxy_server.py b/litellm/proxy/proxy_server.py
index 2343fe8c35d..16a78da4436 100644
--- a/litellm/proxy/proxy_server.py
+++ b/litellm/proxy/proxy_server.py
@@ -11598,3 +11598,5 @@ async def dynamic_mcp_route(mcp_server_name: str, request: Request):
 app.mount(path=BASE_MCP_ROUTE, app=mcp_app)
 app.include_router(mcp_rest_endpoints_router)
 app.include_router(mcp_discoverable_endpoints_router)
+
+# Mock chat completions API (for testing) is started once on port 8090 from proxy_cli._start_mock_chat_api_once()