Update app.py

app.py

@@ -19,6 +19,7 @@ from llama_cpp_agent.tools import WebSearchTool
 from llama_cpp_agent.prompt_templates import web_search_system_prompt, research_system_prompt
 from langchain_community.llms import HuggingFaceHub
 from llama_cpp_agent.llm_output_settings import LlmStructuredOutputSettings, LlmStructuredOutputType
+from pydantic import BaseModel
 from llama_cpp_agent.llm_output_settings import LlmStructuredOutputType
 
 print("Available LlmStructuredOutputType options:")
@@ -100,6 +101,12 @@ examples = [
 ]
 
 
+class CustomLLMSettings(BaseModel):
+    structured_output: LlmStructuredOutputSettings
+    temperature: float
+    top_p: float
+    repetition_penalty: float
+    
 class HuggingFaceHubWrapper:
     def __init__(self, repo_id, model_kwargs, huggingfacehub_api_token):
         self.model = HuggingFaceHub(
@@ -107,13 +114,22 @@ class HuggingFaceHubWrapper:
             model_kwargs=model_kwargs,
             huggingfacehub_api_token=huggingfacehub_api_token
         )
+        self.temperature = model_kwargs.get('temperature', 0.7)
+        self.top_p = model_kwargs.get('top_p', 0.95)
+        self.repetition_penalty = model_kwargs.get('repetition_penalty', 1.1)
+
 
     def get_provider_default_settings(self):
-        return LlmStructuredOutputSettings(
-            output_type=LlmStructuredOutputType.no_structured_output,
-            include_system_prompt=False,
-            include_user_prompt=False,
-            include_assistant_prompt=False,
+        return CustomLLMSettings(
+            structured_output=LlmStructuredOutputSettings(
+                output_type=LlmStructuredOutputType.no_structured_output,
+                include_system_prompt=False,
+                include_user_prompt=False,
+                include_assistant_prompt=False,
+            ),
+            temperature=self.temperature,
+            top_p=self.top_p,
+            repetition_penalty=self.repetition_penalty
         )
 
     def get_provider_identifier(self):
@@ -122,7 +138,6 @@ class HuggingFaceHubWrapper:
     def __call__(self, *args, **kwargs):
         return self.model(*args, **kwargs)
 
-    # Add any other methods that might be required by WebSearchTool
     def get_num_tokens(self, text):
         # This is a placeholder. You might need to implement a proper token counting method
         return len(text.split())