fix(llms/vllm): honor VLLM_BASE_URL instead of always using localhost (#6897)

mem0/configs/llms/vllm.py

@@ -1,3 +1,4 @@
+import os
 from typing import Optional
 
 from mem0.configs.llms.base import BaseLlmConfig
@@ -53,4 +54,4 @@ class VllmConfig(BaseLlmConfig):
         )
 
         # vLLM-specific parameters
-        self.vllm_base_url = vllm_base_url or "http://localhost:8000/v1"
+        self.vllm_base_url = vllm_base_url or os.getenv("VLLM_BASE_URL") or "http://localhost:8000/v1"

mem0/llms/vllm.py

@@ -37,8 +37,7 @@ class VllmLLM(LLMBase):
             self.config.model = "Qwen/Qwen2.5-32B-Instruct"
 
         self.config.api_key = self.config.api_key or os.getenv("VLLM_API_KEY") or "vllm-api-key"
-        base_url = self.config.vllm_base_url or os.getenv("VLLM_BASE_URL")
+        self.client = OpenAI(api_key=self.config.api_key, base_url=self.config.vllm_base_url)
-        self.client = OpenAI(api_key=self.config.api_key, base_url=base_url)
 
     def _parse_response(self, response, tools):
         """

tests/llms/test_vllm.py

@@ -4,6 +4,7 @@ import pytest
 
 from mem0 import AsyncMemory, Memory
 from mem0.configs.llms.base import BaseLlmConfig
+from mem0.configs.llms.vllm import VllmConfig
 from mem0.llms.vllm import VllmLLM
 
 
@@ -15,6 +16,59 @@ def mock_vllm_client():
         yield mock_client
 
 
+def test_base_url_explicit_arg_wins_over_env(mock_vllm_client, monkeypatch):
+    monkeypatch.setenv("VLLM_BASE_URL", "http://from-env:8000/v1")
+    config = VllmConfig(vllm_base_url="http://from-arg:8000/v1")
+    VllmLLM(config)
+
+    assert config.vllm_base_url == "http://from-arg:8000/v1"
+
+
+def test_base_url_env_wins_over_default(mock_vllm_client, monkeypatch):
+    monkeypatch.setenv("VLLM_BASE_URL", "http://from-env:8000/v1")
+    config = VllmConfig()
+    VllmLLM(config)
+
+    assert config.vllm_base_url == "http://from-env:8000/v1"
+
+
+def test_base_url_default_when_neither_set(mock_vllm_client, monkeypatch):
+    monkeypatch.delenv("VLLM_BASE_URL", raising=False)
+    config = VllmConfig()
+    VllmLLM(config)
+
+    assert config.vllm_base_url == "http://localhost:8000/v1"
+
+
+def test_openai_client_receives_explicit_base_url(monkeypatch):
+    monkeypatch.delenv("VLLM_BASE_URL", raising=False)
+    with patch("mem0.llms.vllm.OpenAI") as mock_openai:
+        VllmLLM(VllmConfig(vllm_base_url="http://from-arg:8000/v1"))
+
+    assert mock_openai.call_args.kwargs["base_url"] == "http://from-arg:8000/v1"
+
+
+def test_openai_client_receives_base_url_from_env(monkeypatch):
+    monkeypatch.setenv("VLLM_BASE_URL", "http://from-env:8000/v1")
+    with patch("mem0.llms.vllm.OpenAI") as mock_openai:
+        VllmLLM(VllmConfig())
+
+    base_url = mock_openai.call_args.kwargs["base_url"]
+    assert base_url == "http://from-env:8000/v1"
+    assert base_url != "http://localhost:8000/v1"
+
+
+def test_openai_client_receives_base_url_from_env_via_base_llm_config(monkeypatch):
+    """Original #6256 repro: BaseLlmConfig (not VllmConfig) plus VLLM_BASE_URL only."""
+    monkeypatch.setenv("VLLM_BASE_URL", "http://remote-vllm:8000/v1")
+    with patch("mem0.llms.vllm.OpenAI") as mock_openai:
+        VllmLLM(BaseLlmConfig(model="Qwen/Qwen2.5-32B-Instruct"))
+
+    base_url = mock_openai.call_args.kwargs["base_url"]
+    assert base_url == "http://remote-vllm:8000/v1"
+    assert base_url != "http://localhost:8000/v1"
+
+
 def test_generate_response_without_tools(mock_vllm_client):
     config = BaseLlmConfig(model="Qwen/Qwen2.5-32B-Instruct", temperature=0.7, max_tokens=100, top_p=1.0)
     llm = VllmLLM(config)