Fix Google Gemini model integration with LiteLLM

config/config.yaml

@@ -84,7 +84,7 @@ models:
 
   gemini-2.0-flash-lite:
     provider: "google"
-    model_name: "google/gemini-2.0-flash-lite-001"
+    model_name: "gemini-2.0-flash-lite-001"
     temperature: 0.5
     max_tokens: 2048
     top_p: 1.0

query/llm_interface.py

@@ -101,7 +101,7 @@ class LLMInterface:
                 }
         elif provider == 'google':
             # Special handling for Google Gemini models
-            params['model'] = self.model_config.get('model_name', self.model_name)
+            params['model'] = f"google/{self.model_config.get('model_name', self.model_name)}"
             # Google Gemini uses a different API base
             params['api_base'] = self.model_config.get('endpoint', 'https://generativelanguage.googleapis.com/v1')
         else:

report/report_synthesis.py

@@ -95,6 +95,11 @@ class ReportSynthesizer:
                 'HTTP-Referer': 'https://sim-search.app',  # Replace with your actual app URL
                 'X-Title': 'Intelligent Research System'   # Replace with your actual app name
             }
+        elif provider == 'google':
+            # Special handling for Google Gemini models
+            params['model'] = f"google/{self.model_config.get('model_name', self.model_name)}"
+            # Google Gemini uses a different API base
+            params['api_base'] = self.model_config.get('endpoint', 'https://generativelanguage.googleapis.com/v1')
         else:
             # Standard provider (OpenAI, Anthropic, etc.)
             params['model'] = self.model_name
@@ -118,6 +123,9 @@ class ReportSynthesizer:
             params['messages'] = messages
             params['stream'] = stream
             
+            logger.info(f"Generating completion with model: {params.get('model')}")
+            logger.info(f"Provider: {self.model_config.get('provider')}")
+            
             response = completion(**params)
             
             if stream:
@@ -132,6 +140,13 @@ class ReportSynthesizer:
                 return content
         except Exception as e:
             logger.error(f"Error generating completion: {e}")
+            logger.error(f"Model params: {params}")
+            
+            # More detailed error for debugging
+            if hasattr(e, '__dict__'):
+                for key, value in e.__dict__.items():
+                    logger.error(f"Error detail - {key}: {value}")
+            
             return f"Error: {str(e)}"
     
     def _process_thinking_tags(self, content: str) -> str: