fix(openai-compatible): support custom temperature setting (#60)

notebook_intelligence/llm_providers/openai_compatible_llm_provider.py

@@ -10,6 +10,21 @@
 
 DEFAULT_CONTEXT_WINDOW = 4096
 
+
+def _parse_temperature(prop: LLMProviderProperty | None):
+    if prop is None:
+        return omit
+
+    value = prop.value
+    if value is None:
+        return omit
+
+    value = str(value).strip()
+    if value == "":
+        return omit
+
+    return float(value)
+
 class OpenAICompatibleChatModel(ChatModel):
     def __init__(self, provider: "OpenAICompatibleLLMProvider"):
         super().__init__(provider)
@@ -19,6 +34,7 @@ def __init__(self, provider: "OpenAICompatibleLLMProvider"):
             LLMProviderProperty("model_id", "Model", "Model (must support streaming)", "", False),
             LLMProviderProperty("base_url", "Base URL", "Base URL", "", True),
             LLMProviderProperty("context_window", "Context window", "Context window length", "", True),
+            LLMProviderProperty("temperature", "Temperature", "Sampling temperature", "", True),
         ]
 
     @property
@@ -46,13 +62,15 @@ def completions(self, messages: list[dict], tools: list[dict] = None, response:
         base_url = base_url_prop.value if base_url_prop is not None else None
         base_url = base_url if base_url.strip() != "" else None
         api_key = self.get_property("api_key").value
+        temperature = _parse_temperature(self.get_property("temperature"))
 
         client = OpenAI(base_url=base_url, api_key=api_key)
         resp = client.chat.completions.create(
             model=model_id,
             messages=messages.copy(),
             tools=tools or omit,
             tool_choice=options.get("tool_choice", omit),
+            temperature=temperature,
             stream=stream,
         )
 
@@ -94,6 +112,7 @@ def __init__(self, provider: "OpenAICompatibleLLMProvider"):
             LLMProviderProperty("model_id", "Model", "Model", "", False),
             LLMProviderProperty("base_url", "Base URL", "Base URL", "", True),
             LLMProviderProperty("context_window", "Context window", "Context window length", "", True),
+            LLMProviderProperty("temperature", "Temperature", "Sampling temperature", "", True),
         ]
 
     @property
@@ -142,6 +161,7 @@ def inline_completions(self, prefix, suffix, language, filename, context: Comple
         base_url = base_url_prop.value if base_url_prop is not None else None
         base_url = base_url if base_url and base_url.strip() != "" else None
         api_key = self.get_property("api_key").value
+        temperature = _parse_temperature(self.get_property("temperature"))
 
         client = OpenAI(base_url=base_url, api_key=api_key)
         resp = client.chat.completions.create(
@@ -154,6 +174,7 @@ def inline_completions(self, prefix, suffix, language, filename, context: Comple
 """}
             ],
             max_tokens=1000,
+            temperature=temperature,
             stream=False,
         )
 

tests/test_openai_compatible_llm_provider.py

@@ -0,0 +1,62 @@
+from unittest.mock import Mock, patch
+
+from openai import omit
+
+from notebook_intelligence.llm_providers.openai_compatible_llm_provider import (
+    OpenAICompatibleLLMProvider,
+    _parse_temperature,
+)
+from notebook_intelligence.api import LLMProviderProperty
+
+
+def test_parse_temperature_omits_missing_values():
+    assert _parse_temperature(None) is omit
+    assert _parse_temperature(LLMProviderProperty("temperature", "Temperature", "", "   ", True)) is omit
+
+
+def test_parse_temperature_returns_float_value():
+    prop = LLMProviderProperty("temperature", "Temperature", "", "0.25", True)
+    assert _parse_temperature(prop) == 0.25
+
+
+@patch("notebook_intelligence.llm_providers.openai_compatible_llm_provider.OpenAI")
+def test_chat_completions_passes_temperature(mock_openai_cls):
+    provider = OpenAICompatibleLLMProvider()
+    model = provider.chat_models[0]
+    model.set_property_value("api_key", "test-key")
+    model.set_property_value("model_id", "gpt-4.1")
+    model.set_property_value("temperature", "0.2")
+
+    mock_client = Mock()
+    mock_response = Mock()
+    mock_response.model_dump_json.return_value = '{"choices": []}'
+    mock_response.choices = []
+    mock_client.chat.completions.create.return_value = mock_response
+    mock_openai_cls.return_value = mock_client
+
+    model.completions(messages=[{"role": "user", "content": "hi"}])
+
+    mock_client.chat.completions.create.assert_called_once()
+    assert mock_client.chat.completions.create.call_args.kwargs["temperature"] == 0.2
+
+
+@patch("notebook_intelligence.llm_providers.openai_compatible_llm_provider.OpenAI")
+def test_inline_completions_omits_blank_temperature(mock_openai_cls):
+    provider = OpenAICompatibleLLMProvider()
+    model = provider.inline_completion_models[0]
+    model.set_property_value("api_key", "test-key")
+    model.set_property_value("model_id", "gpt-4.1")
+    model.set_property_value("temperature", "")
+
+    mock_client = Mock()
+    mock_response = Mock()
+    mock_response.choices = [Mock(message=Mock(content="```python\npass\n```"))]
+    mock_client.chat.completions.create.return_value = mock_response
+    mock_openai_cls.return_value = mock_client
+
+    cancel_token = Mock(is_cancel_requested=False)
+    result = model.inline_completions("", "", "python", "test.py", Mock(), cancel_token)
+
+    assert result.strip() == "pass"
+    mock_client.chat.completions.create.assert_called_once()
+    assert mock_client.chat.completions.create.call_args.kwargs["temperature"] is omit