Use better, standard default temp, top_p for openai model providers

src/khoj/processor/conversation/openai/gpt.py

@@ -85,10 +85,10 @@ async def converse_openai(
     program_execution_context: List[str] = None,
     location_data: LocationData = None,
     chat_history: list[ChatMessageModel] = [],
-    model: str = "gpt-4o-mini",
+    model: str = "gpt-4.1-mini",
     api_key: Optional[str] = None,
     api_base_url: Optional[str] = None,
-    temperature: float = 0.4,
+    temperature: float = 0.6,
     max_prompt_size=None,
     tokenizer_name=None,
     user_name: str = None,

src/khoj/processor/conversation/openai/utils.py

@@ -71,7 +71,7 @@ openai_async_clients: Dict[str, openai.AsyncOpenAI] = {}
 def completion_with_backoff(
     messages: List[ChatMessage],
     model_name: str,
-    temperature=0.8,
+    temperature=0.6,
     openai_api_key=None,
     api_base_url=None,
     deepthought: bool = False,
@@ -89,14 +89,19 @@ def completion_with_backoff(
     if stream:
         model_kwargs["stream_options"] = {"include_usage": True}
 
+    model_kwargs["temperature"] = temperature
+    model_kwargs["top_p"] = model_kwargs.get("top_p", 0.95)
+
     formatted_messages = format_message_for_api(messages, api_base_url)
 
     # Tune reasoning models arguments
     if is_openai_reasoning_model(model_name, api_base_url):
-        temperature = 1
+        model_kwargs["temperature"] = 1
         reasoning_effort = "medium" if deepthought else "low"
         model_kwargs["reasoning_effort"] = reasoning_effort
+        model_kwargs.pop("top_p", None)
     elif is_twitter_reasoning_model(model_name, api_base_url):
+        model_kwargs.pop("temperature", None)
         reasoning_effort = "high" if deepthought else "low"
         model_kwargs["reasoning_effort"] = reasoning_effort
     elif model_name.startswith("deepseek-reasoner"):
@@ -131,7 +136,6 @@ def completion_with_backoff(
         with client.beta.chat.completions.stream(
             messages=formatted_messages,  # type: ignore
             model=model_name,
-            temperature=temperature,
             timeout=httpx.Timeout(30, read=read_timeout),
             **model_kwargs,
         ) as chat:
@@ -233,9 +237,7 @@ async def chat_completion_with_backoff(
     openai_api_key=None,
     api_base_url=None,
     deepthought=False,
-    model_kwargs: dict = {},
     tracer: dict = {},
-    tools=None,
 ) -> AsyncGenerator[ResponseWithThought, None]:
     client_key = f"{openai_api_key}--{api_base_url}"
     client = openai_async_clients.get(client_key)
@@ -243,6 +245,7 @@ async def chat_completion_with_backoff(
         client = get_openai_async_client(openai_api_key, api_base_url)
         openai_async_clients[client_key] = client
 
+    model_kwargs: dict = {}
     stream = not is_non_streaming_model(model_name, api_base_url)
     stream_processor = astream_thought_processor
     if stream:
@@ -250,6 +253,8 @@ async def chat_completion_with_backoff(
     else:
         model_kwargs.pop("stream_options", None)
 
+    model_kwargs["top_p"] = model_kwargs.get("top_p", 0.95)
+
     formatted_messages = format_message_for_api(messages, api_base_url)
 
     # Configure thinking for openai reasoning models
@@ -257,7 +262,9 @@ async def chat_completion_with_backoff(
         temperature = 1
         reasoning_effort = "medium" if deepthought else "low"
         model_kwargs["reasoning_effort"] = reasoning_effort
-        model_kwargs.pop("stop", None)  # Remove unsupported stop param for reasoning models
+        # Remove unsupported params for reasoning models
+        model_kwargs.pop("top_p", None)
+        model_kwargs.pop("stop", None)
 
         # Get the first system message and add the string `Formatting re-enabled` to it.
         # See https://platform.openai.com/docs/guides/reasoning-best-practices
@@ -304,8 +311,6 @@ async def chat_completion_with_backoff(
     read_timeout = 300 if is_local_api(api_base_url) else 60
     if os.getenv("KHOJ_LLM_SEED"):
         model_kwargs["seed"] = int(os.getenv("KHOJ_LLM_SEED"))
-    if tools:
-        model_kwargs["tools"] = tools
 
     aggregated_response = ""
     final_chunk = None