Add OPENAI_LLM_TIMEOUT env var for custom OpenAI-compatible provider

hermes_cli/config.py

@@ -569,6 +569,15 @@ def ensure_hermes_home():
         "url": None,
         "password": False,
         "category": "setting",
+
+    },
+    "OPENAI_LLM_TIMEOUT": {
+        "description": "Timeout in seconds for LLM calls when using custom OpenAI-compatible provider",
+        "prompt": "LLM timeout for custom provider",
+        "url": None,
+        "password": False,
+        "category": "setting",
+        "advanced": True,
     },
 }
 

run_agent.py

@@ -2390,11 +2390,19 @@ def _build_api_kwargs(self, api_messages: list) -> dict:
         if self.provider_data_collection:
             provider_preferences["data_collection"] = self.provider_data_collection
 
+        # Determine timeout based on provider and environment variable
+        timeout = 900.0
+        if self.provider == "custom" and os.getenv("OPENAI_LLM_TIMEOUT"):
+            try:
+                timeout = float(os.getenv("OPENAI_LLM_TIMEOUT"))
+            except ValueError:
+                pass  # Keep default if invalid
+
         api_kwargs = {
             "model": self.model,
             "messages": api_messages,
             "tools": self.tools if self.tools else None,
-            "timeout": 900.0,
+            "timeout": timeout,
         }
 
         if self.max_tokens is not None:

tests/test_run_agent.py

@@ -490,12 +490,48 @@ def test_basic_kwargs(self, agent):
         assert kwargs["messages"] is messages
         assert kwargs["timeout"] == 900.0
 
+    def test_timeout_custom_provider_with_env(self, agent, monkeypatch):
+        import os
+
+        monkeypatch.setenv("OPENAI_LLM_TIMEOUT", "120.0")
+        agent.provider = "custom"
+        messages = [{"role": "user", "content": "hi"}]
+        kwargs = agent._build_api_kwargs(messages)
+        assert kwargs["timeout"] == 120.0
+
+    def test_timeout_custom_provider_without_env(self, agent):
+        import os
+
+        os.environ.pop("OPENAI_LLM_TIMEOUT", None)
+        agent.provider = "custom"
+        messages = [{"role": "user", "content": "hi"}]
+        kwargs = agent._build_api_kwargs(messages)
+        assert kwargs["timeout"] == 900.0
+
+    def test_timeout_openrouter_ignores_env(self, agent, monkeypatch):
+        import os
+
+        monkeypatch.setenv("OPENAI_LLM_TIMEOUT", "120.0")
+        agent.provider = "openrouter"
+        messages = [{"role": "user", "content": "hi"}]
+        kwargs = agent._build_api_kwargs(messages)
+        assert kwargs["timeout"] == 900.0
+
+    def test_timeout_invalid_env_value(self, agent, monkeypatch):
+        import os
+
+        monkeypatch.setenv("OPENAI_LLM_TIMEOUT", "invalid")
+        agent.provider = "custom"
+        messages = [{"role": "user", "content": "hi"}]
+        kwargs = agent._build_api_kwargs(messages)
+        assert kwargs["timeout"] == 900.0
+
     def test_provider_preferences_injected(self, agent):
         agent.providers_allowed = ["Anthropic"]
         messages = [{"role": "user", "content": "hi"}]
         kwargs = agent._build_api_kwargs(messages)
+        assert "provider" in kwargs.get("extra_body", {})
         assert kwargs["extra_body"]["provider"]["only"] == ["Anthropic"]
-
     def test_reasoning_config_default_openrouter(self, agent):
         """Default reasoning config for OpenRouter should be medium."""
         messages = [{"role": "user", "content": "hi"}]