fix(llm): pass the output-token cap and a configured endpoint to Azure OpenAI (#1584, #1600)

- max_tokens reaches the Azure deployment
- a tier's backend_url is used as the Azure endpoint, else AZURE_OPENAI_ENDPOINT

tests/test_azure_client.py

@@ -0,0 +1,28 @@
+"""Azure OpenAI takes the run's settings like every other provider."""
+
+from __future__ import annotations
+
+import pytest
+
+from tradingagents.llm_clients.factory import create_llm_client
+
+
+@pytest.fixture(autouse=True)
+def _azure_env(monkeypatch):
+    monkeypatch.setenv("AZURE_OPENAI_API_KEY", "test-key")
+    monkeypatch.setenv("AZURE_OPENAI_ENDPOINT", "https://from-env.openai.azure.com/")
+    monkeypatch.setenv("OPENAI_API_VERSION", "2025-03-01-preview")
+    monkeypatch.delenv("AZURE_OPENAI_DEPLOYMENT_NAME", raising=False)
+
+
+@pytest.mark.unit
+def test_the_output_token_cap_reaches_the_deployment():
+    llm = create_llm_client("azure", "gpt-5.5", max_tokens=1234).get_llm()
+    assert llm.max_tokens == 1234
+
+
+@pytest.mark.unit
+def test_an_explicit_endpoint_wins_over_the_environment():
+    llm = create_llm_client("azure", "gpt-5.5", base_url="https://deep.openai.azure.com/").get_llm()
+    assert llm.azure_endpoint == "https://deep.openai.azure.com/"
+    assert create_llm_client("azure", "gpt-5.5").get_llm().azure_endpoint == "https://from-env.openai.azure.com/"

tradingagents/llm_clients/azure_client.py

@@ -6,7 +6,7 @@ from langchain_openai import AzureChatOpenAI
 from .base_client import BaseLLMClient, normalize_content
 
 _PASSTHROUGH_KWARGS = (
-    "timeout", "max_retries", "api_key", "reasoning_effort", "temperature",
+    "timeout", "max_retries", "api_key", "reasoning_effort", "temperature", "max_tokens",
     "callbacks", "http_client", "http_async_client",
 )
 
@@ -23,7 +23,8 @@ class AzureOpenAIClient(BaseLLMClient):
 
     Requires environment variables:
         AZURE_OPENAI_API_KEY: API key
-        AZURE_OPENAI_ENDPOINT: Endpoint URL (e.g. https://<resource>.openai.azure.com/)
+        AZURE_OPENAI_ENDPOINT: Endpoint URL (e.g. https://<resource>.openai.azure.com/),
+            unless backend_url gives one
         AZURE_OPENAI_DEPLOYMENT_NAME: Deployment name
         OPENAI_API_VERSION: API version (e.g. 2025-03-01-preview)
     """
@@ -39,6 +40,8 @@ class AzureOpenAIClient(BaseLLMClient):
             "model": self.model,
             "azure_deployment": os.environ.get("AZURE_OPENAI_DEPLOYMENT_NAME", self.model),
         }
+        if self.base_url:   # a configured endpoint, else AZURE_OPENAI_ENDPOINT
+            llm_kwargs["azure_endpoint"] = self.base_url
 
         for key in _PASSTHROUGH_KWARGS:
             if key in self.kwargs: