Fix model provider selection in report generation to ensure correct provider is used based on config

report/report_synthesis.py

@@ -96,7 +96,25 @@ class ReportSynthesizer:
     
     def _setup_provider(self) -> None:
         """Set up the LLM provider based on the model configuration."""
-        provider = self.model_config.get('provider', 'groq')
+        # Determine provider based on model name if not explicitly specified
+        if 'provider' in self.model_config:
+            provider = self.model_config['provider']
+        elif 'gemini' in self.model_name.lower():
+            provider = 'gemini'
+            # Update model_config to reflect the correct provider
+            self.model_config['provider'] = 'gemini'
+        elif 'claude' in self.model_name.lower():
+            provider = 'anthropic'
+            self.model_config['provider'] = 'anthropic'
+        elif 'gpt' in self.model_name.lower() or 'text-embedding' in self.model_name.lower():
+            provider = 'openai'
+            self.model_config['provider'] = 'openai'
+        elif 'mistral' in self.model_name.lower():
+            provider = 'mistral'
+            self.model_config['provider'] = 'mistral'
+        else:
+            provider = 'groq'  # Default to groq for other models
+            self.model_config['provider'] = 'groq'
         
         # Log detailed model information for debugging
         logger.info(f"Setting up report synthesizer with model: {self.model_name} (provider: {provider})")
@@ -207,9 +225,12 @@ class ReportSynthesizer:
             If stream is False, returns the completion text as a string
             If stream is True, returns the completion response object for streaming
         """
-        # Get provider from model config
+        # Get provider from model config - this should be set correctly from the config file
         provider = self.model_config.get('provider', 'groq').lower()
         
+        # Log the provider and model being used for debugging
+        logger.info(f"Using provider: {provider} with model: {self.model_name}")
+        
         # Special handling for Gemini models - they use 'user' and 'model' roles
         if provider == 'gemini':
             formatted_messages = []
@@ -236,11 +257,29 @@ class ReportSynthesizer:
         # Get completion parameters
         params = self._get_completion_params()
         
-        # Double-check that we're using the correct model
+        # Use the provider from the model config - this should be set correctly from the config file
+        # Different providers have different formatting requirements for the model parameter
         if provider == 'gemini':
-            params['model'] = f"gemini/{self.model_name}"
+            # For Gemini models, we use the vertex_ai provider in litellm
+            params['model'] = self.model_name
+            params['custom_llm_provider'] = 'vertex_ai'
         elif provider == 'groq':
-            params['model'] = f"groq/{self.model_name}"
+            # For Groq models, we need to prefix with provider
+            if not self.model_name.startswith('groq/'):
+                params['model'] = f"groq/{self.model_name}"
+            else:
+                params['model'] = self.model_name
+        elif provider == 'anthropic':
+            # For Claude models
+            params['model'] = self.model_name
+            params['custom_llm_provider'] = 'anthropic'
+        elif provider == 'openai':
+            # For OpenAI models
+            params['model'] = self.model_name
+            params['custom_llm_provider'] = 'openai'
+        
+        # Log the final model parameter being used
+        logger.info(f"Final model parameter: {params.get('model', 'unknown')} with provider: {provider}")
         
         # Log the actual parameters being used for the LLM call
         safe_params = params.copy()