From 19dae36f581cff030dd15b2d94a9b54261229065 Mon Sep 17 00:00:00 2001
From: Charlie Tonneslan <cst0520@gmail.com>
Date: Sun, 22 Mar 2026 12:47:32 -0400
Subject: [PATCH] Pass model name to _get_provider_kwargs for per-model
 decisions

The previous fix only checked deep_think_llm when deciding whether
to include reasoning_effort, but both LLMs shared the same kwargs.
This meant mixing an o-series deep model with a gpt-* quick model
would still crash.

Now each LLM gets its own kwargs by passing the model name to
_get_provider_kwargs. reasoning_effort is only added when the
specific model being configured is an o-series model.
---
 tradingagents/graph/trading_graph.py | 33 ++++++++++++++++++----------
 1 file changed, 22 insertions(+), 11 deletions(-)

diff --git a/tradingagents/graph/trading_graph.py b/tradingagents/graph/trading_graph.py
index 1116b003..bd0c578b 100644
--- a/tradingagents/graph/trading_graph.py
+++ b/tradingagents/graph/trading_graph.py
@@ -71,24 +71,31 @@ class TradingAgentsGraph:
             exist_ok=True,
         )
 
-        # Initialize LLMs with provider-specific thinking configuration
-        llm_kwargs = self._get_provider_kwargs()
+        # Initialize LLMs with provider-specific thinking configuration.
+        # Each model gets its own kwargs since options like reasoning_effort
+        # are only valid for certain model families.
+        deep_model = self.config["deep_think_llm"]
+        quick_model = self.config["quick_think_llm"]
+
+        deep_kwargs = self._get_provider_kwargs(deep_model)
+        quick_kwargs = self._get_provider_kwargs(quick_model)
 
         # Add callbacks to kwargs if provided (passed to LLM constructor)
         if self.callbacks:
-            llm_kwargs["callbacks"] = self.callbacks
+            deep_kwargs["callbacks"] = self.callbacks
+            quick_kwargs["callbacks"] = self.callbacks
 
         deep_client = create_llm_client(
             provider=self.config["llm_provider"],
-            model=self.config["deep_think_llm"],
+            model=deep_model,
             base_url=self.config.get("backend_url"),
-            **llm_kwargs,
+            **deep_kwargs,
         )
         quick_client = create_llm_client(
             provider=self.config["llm_provider"],
-            model=self.config["quick_think_llm"],
+            model=quick_model,
             base_url=self.config.get("backend_url"),
-            **llm_kwargs,
+            **quick_kwargs,
         )
 
         self.deep_thinking_llm = deep_client.get_llm()
@@ -133,8 +140,13 @@ class TradingAgentsGraph:
         # Set up the graph
         self.graph = self.graph_setup.setup_graph(selected_analysts)
 
-    def _get_provider_kwargs(self) -> Dict[str, Any]:
-        """Get provider-specific kwargs for LLM client creation."""
+    def _get_provider_kwargs(self, model_name: str) -> Dict[str, Any]:
+        """Get provider-specific kwargs for LLM client creation.
+
+        Args:
+            model_name: The model being configured, used to decide which
+                        provider-specific options are applicable.
+        """
         kwargs = {}
         provider = self.config.get("llm_provider", "").lower()
 
@@ -148,8 +160,7 @@ class TradingAgentsGraph:
             if reasoning_effort:
                 # reasoning_effort is only supported by o-series models
                 # (o1, o3, o3-mini, o4-mini, etc.), not by gpt-* models
-                model = self.config.get("deep_think_llm", "")
-                if model.startswith("o"):
+                if model_name.startswith("o"):
                     kwargs["reasoning_effort"] = reasoning_effort
 
         return kwargs