add llama

codes/llm_provider.py

@@ -79,7 +79,7 @@ class OpenAIProvider(LLMProvider):
 
 
 class GeminiProvider(LLMProvider):
-    """Google Gemini API implementat`ion"""
+    """Google Gemini API implementation"""
     
     def __init__(self, api_key: Optional[str] = None):
         try:
@@ -94,18 +94,13 @@ class GeminiProvider(LLMProvider):
     
     def create_completion(self, messages: List[Dict], model: str, **kwargs) -> Any:
         """Create Gemini chat completion"""
-        # Convert OpenAI message format to Gemini format
         gemini_messages = self._convert_messages(messages)
-        
         # Do NOT add models/ prefix - pass model name directly
         gemini_model = self.genai.GenerativeModel(model)
-        
-        # Generate response
         response = gemini_model.generate_content(
             gemini_messages,
             generation_config=self._get_generation_config(**kwargs)
         )
-        
         return response
     
     def _convert_messages(self, messages: List[Dict]) -> str:
@@ -175,7 +170,7 @@ class GeminiProvider(LLMProvider):
 
 
 class GemmaProvider(LLMProvider):
-    """NVIDIA Gemma API implementation"""
+    """NVIDIA API implementation — supports Gemma, Llama, and other NVIDIA-hosted models"""
     
     def __init__(self, api_key: Optional[str] = None):
         import requests
@@ -189,7 +184,7 @@ class GemmaProvider(LLMProvider):
         self.invoke_url = "https://integrate.api.nvidia.com/v1/chat/completions"
     
     def create_completion(self, messages: List[Dict], model: str, **kwargs) -> Any:
-        """Create Gemma chat completion"""
+        """Create NVIDIA API chat completion"""
         headers = {
             "Authorization": f"Bearer {self.api_key}",
             "Accept": "application/json"
@@ -197,7 +192,7 @@ class GemmaProvider(LLMProvider):
         payload = {
             "model": model,
             "messages": messages,
-            "max_tokens": kwargs.get('max_tokens', 512),
+            "max_tokens": kwargs.get('max_tokens', 8192),  # increased for code generation
             "temperature": kwargs.get('temperature', 0.20),
             "top_p": kwargs.get('top_p', 0.70),
             "stream": False
@@ -207,13 +202,13 @@ class GemmaProvider(LLMProvider):
         return response.json()
     
     def get_response_text(self, completion: Any) -> str:
-        """Extract text from Gemma response"""
+        """Extract text from NVIDIA API response"""
         if isinstance(completion, dict):
             return completion['choices'][0]['message']['content']
         return str(completion)
     
     def get_usage_info(self, completion: Any) -> Dict:
-        """Extract usage from Gemma response"""
+        """Extract usage from NVIDIA API response"""
         try:
             usage = completion.get('usage', {})
             return {
@@ -231,9 +226,11 @@ class GemmaProvider(LLMProvider):
             }
     
     def calculate_cost(self, usage: Dict, model: str) -> float:
-        """Calculate Gemma cost"""
+        """Calculate NVIDIA API cost"""
         model_costs = {
             "google/gemma-3-27b-it": {"input": 0.0, "output": 0.0},
+            "meta/llama-3.3-70b-instruct": {"input": 0.0, "output": 0.0},
+            "meta/llama-3.1-8b-instruct": {"input": 0.0, "output": 0.0},
         }
         costs = model_costs.get(model, {"input": 0.0, "output": 0.0})
         prompt_tokens = usage['prompt_tokens']
@@ -262,8 +259,8 @@ def get_default_model(provider_name: str) -> str:
     """Get default model for a provider"""
     defaults = {
         'openai': 'gpt-4o-mini',
-        'gemini': 'gemini-2.0-flash',   # valid, free tier, no models/ prefix needed
-        'gemma': 'google/gemma-3-27b-it',
+        'gemini': 'gemini-1.5-flash',
+        'gemma': 'meta/llama-3.3-70b-instruct',  # Llama via NVIDIA API
     }
     return defaults.get(provider_name, 'gpt-4o-mini')