From 59b0337ff006d00f87579f8b65077726404d022a Mon Sep 17 00:00:00 2001
From: Dave Fobare <dave.fobare@gmail.com>
Date: Fri, 15 May 2026 16:27:40 -0400
Subject: [PATCH] Respect explicit Ollama think setting

---
 SourceCode/shared_tools/ollama_client.py |  2 +-
 tests/test_ollama_wait_for_available.py  | 46 ++++++++++++++++++++++++
 2 files changed, 47 insertions(+), 1 deletion(-)

diff --git a/SourceCode/shared_tools/ollama_client.py b/SourceCode/shared_tools/ollama_client.py
index fc0f679..31300b9 100755
--- a/SourceCode/shared_tools/ollama_client.py
+++ b/SourceCode/shared_tools/ollama_client.py
@@ -202,7 +202,7 @@ def chat(
                 },
             }
             effective_think = think
-            if self._is_reasoning_model(model_name):
+            if effective_think is None and self._is_reasoning_model(model_name):
                 effective_think = True
             if effective_think is not None:
                 payload["think"] = effective_think
diff --git a/tests/test_ollama_wait_for_available.py b/tests/test_ollama_wait_for_available.py
index 45af9ac..1d7fca9 100644
--- a/tests/test_ollama_wait_for_available.py
+++ b/tests/test_ollama_wait_for_available.py
@@ -8,6 +8,52 @@
 
 
 class OllamaWaitForAvailableTests(unittest.TestCase):
+    def test_chat_auto_enables_thinking_for_reasoning_models_by_default(self) -> None:
+        client = OllamaClient()
+        seen_payloads: list[dict] = []
+
+        def fake_stream(payload: dict, *, timeout: int, idle_timeout_sec: int) -> str:
+            seen_payloads.append(dict(payload))
+            return "ok"
+
+        client._post_json_stream_chat = MagicMock(side_effect=fake_stream)
+
+        out = client.chat("qwen3:8b", "system", "user")
+
+        self.assertEqual(out, "ok")
+        self.assertEqual(seen_payloads[0].get("think"), True)
+
+    def test_chat_respects_explicit_false_think_for_reasoning_models(self) -> None:
+        client = OllamaClient()
+        seen_payloads: list[dict] = []
+
+        def fake_stream(payload: dict, *, timeout: int, idle_timeout_sec: int) -> str:
+            seen_payloads.append(dict(payload))
+            return "ok"
+
+        client._post_json_stream_chat = MagicMock(side_effect=fake_stream)
+
+        out = client.chat("qwen3:14b", "system", "user", think=False, num_predict=48)
+
+        self.assertEqual(out, "ok")
+        self.assertEqual(seen_payloads[0].get("think"), False)
+        self.assertEqual(seen_payloads[0]["options"]["num_predict"], 48)
+
+    def test_chat_respects_explicit_true_think_for_non_reasoning_models(self) -> None:
+        client = OllamaClient()
+        seen_payloads: list[dict] = []
+
+        def fake_stream(payload: dict, *, timeout: int, idle_timeout_sec: int) -> str:
+            seen_payloads.append(dict(payload))
+            return "ok"
+
+        client._post_json_stream_chat = MagicMock(side_effect=fake_stream)
+
+        out = client.chat("qwen2.5-coder:7b", "system", "user", think=True)
+
+        self.assertEqual(out, "ok")
+        self.assertEqual(seen_payloads[0].get("think"), True)
+
     def test_wait_for_available_uses_tags_and_show_not_chat(self) -> None:
         client = OllamaClient()
         client._get_json = MagicMock(return_value={"models": [{"name": "qwen3:8b"}]})