Update app.py

app.py

@@ -639,8 +639,12 @@ def is_vision_model(provider, model_name):
             return True
             
         # Also check for common vision indicators in model names
-        if any(x in model_name.lower() for x in ["vl", "vision", "visual", "llava", "gemini"]):
-            return True
+        try:
+            if any(x in model_name.lower() for x in ["vl", "vision", "visual", "llava", "gemini"]):
+                return True
+        except AttributeError:
+            # In case model_name is not a string or has no lower method
+            return False
     
     return False
 
@@ -806,21 +810,28 @@ def call_groq_api(payload, api_key_override=None):
         
         # Extract parameters from payload
         model = payload.get("model", "llama-3.1-8b-instant")
-        messages = payload.get("messages", [])
-        temperature = payload.get("temperature", 0.7)
-        max_tokens = payload.get("max_tokens", 1000)
-        stream = payload.get("stream", False)
-        top_p = payload.get("top_p", 0.9)
+        
+        # Clean up messages - remove any unexpected properties
+        messages = []
+        for msg in payload.get("messages", []):
+            clean_msg = {
+                "role": msg["role"],
+                "content": msg["content"]
+            }
+            messages.append(clean_msg)
+        
+        # Basic parameters
+        groq_payload = {
+            "model": model,
+            "messages": messages,
+            "temperature": payload.get("temperature", 0.7),
+            "max_tokens": payload.get("max_tokens", 1000),
+            "stream": payload.get("stream", False),
+            "top_p": payload.get("top_p", 0.9)
+        }
         
         # Create completion
-        response = client.chat.completions.create(
-            model=model,
-            messages=messages,
-            temperature=temperature,
-            max_tokens=max_tokens,
-            stream=stream,
-            top_p=top_p
-        )
+        response = client.chat.completions.create(**groq_payload)
         
         return response
     except Exception as e:
@@ -837,7 +848,7 @@ def call_cohere_api(payload, api_key_override=None):
         if not api_key:
             raise ValueError("Cohere API key is required")
             
-        client = cohere.Client(api_key=api_key)
+        client = cohere.ClientV2(api_key=api_key)
         
         # Extract parameters from payload
         model = payload.get("model", "command-r-plus")
@@ -845,40 +856,27 @@ def call_cohere_api(payload, api_key_override=None):
         temperature = payload.get("temperature", 0.7)
         max_tokens = payload.get("max_tokens", 1000)
         
-        # Format messages for Cohere
-        chat_history = []
-        user_message = ""
-        
+        # Transform messages to Cohere format - IMPORTANT 
+        # Cohere uses specific role names: USER, ASSISTANT, SYSTEM, TOOL
+        cohere_messages = []
         for msg in messages:
-            if msg["role"] == "system":
-                # For system message, we'll prepend to the user's first message
-                system_content = msg["content"]
-                if isinstance(system_content, list):  # Handle multimodal content
-                    system_parts = []
-                    for item in system_content:
-                        if item["type"] == "text":
-                            system_parts.append(item["text"])
-                    system_content = "\n".join(system_parts)
-                user_message = f"System: {system_content}\n\n" + user_message
-            elif msg["role"] == "user":
-                content = msg["content"]
-                # Handle multimodal content
-                if isinstance(content, list):
-                    text_parts = []
-                    for item in content:
-                        if item["type"] == "text":
-                            text_parts.append(item["text"])
-                    content = "\n".join(text_parts)
-                user_message = content
-            elif msg["role"] == "assistant":
-                content = msg["content"]
-                if content:
-                    chat_history.append({"role": "ASSISTANT", "message": content})
+            role = msg["role"].upper()  # Cohere requires uppercase roles
+            content = msg["content"]
+            
+            # Handle multimodal content
+            if isinstance(content, list):
+                text_parts = []
+                for item in content:
+                    if item["type"] == "text":
+                        text_parts.append(item["text"])
+                content = "\n".join(text_parts)
+                
+            cohere_messages.append({"role": role, "content": content})
         
         # Create chat completion
         response = client.chat(
-            message=user_message,
-            chat_history=chat_history,
+            message=cohere_messages[-1]["content"] if cohere_messages else "",
+            chat_history=cohere_messages[:-1] if len(cohere_messages) > 1 else [],
             model=model,
             temperature=temperature,
             max_tokens=max_tokens
@@ -898,7 +896,8 @@ def call_together_api(payload, api_key_override=None):
         api_key = api_key_override if api_key_override else TOGETHER_API_KEY
         if not api_key:
             raise ValueError("Together API key is required")
-            
+        
+        # Create client with Together base URL
         client = openai.OpenAI(
             api_key=api_key,
             base_url="https://api.together.xyz/v1"
@@ -906,19 +905,34 @@ def call_together_api(payload, api_key_override=None):
         
         # Extract parameters from payload
         model = payload.get("model", "meta-llama/Llama-3.1-8B-Instruct")
-        messages = payload.get("messages", [])
-        temperature = payload.get("temperature", 0.7)
-        max_tokens = payload.get("max_tokens", 1000)
-        stream = payload.get("stream", False)
+        
+        # Fix model name format - Together API expects this format
+        if not model.startswith("meta-llama/") and "llama" in model.lower():
+            # Convert model ID format from "llama-3.1-8b-instruct" to "meta-llama/Llama-3.1-8B-Instruct"
+            parts = model.split("-")
+            formatted_name = "meta-llama/L" + "".join([p.capitalize() for p in parts])
+            model = formatted_name
+            
+        # Clean up messages - remove any unexpected properties
+        messages = []
+        for msg in payload.get("messages", []):
+            clean_msg = {
+                "role": msg["role"],
+                "content": msg["content"]
+            }
+            messages.append(clean_msg)
+            
+        # Create payload
+        together_payload = {
+            "model": model,
+            "messages": messages,
+            "temperature": payload.get("temperature", 0.7),
+            "max_tokens": payload.get("max_tokens", 1000),
+            "stream": payload.get("stream", False)
+        }
         
         # Create completion
-        response = client.chat.completions.create(
-            model=model,
-            messages=messages,
-            temperature=temperature,
-            max_tokens=max_tokens,
-            stream=stream
-        )
+        response = client.chat.completions.create(**together_payload)
         
         return response
     except Exception as e:
@@ -928,7 +942,7 @@ def call_together_api(payload, api_key_override=None):
 def call_ovh_api(payload, api_key_override=None):
     """Make a call to OVH AI Endpoints API with error handling"""
     try:
-        # Use custom OpenAI client with the OVH endpoint
+        # Extract parameters from payload
         model = payload.get("model", "ovh/llama-3.1-8b-instruct")
         messages = payload.get("messages", [])
         temperature = payload.get("temperature", 0.7)
@@ -938,15 +952,25 @@ def call_ovh_api(payload, api_key_override=None):
             "Content-Type": "application/json"
         }
         
+        # Clean up messages - remove any unexpected properties
+        clean_messages = []
+        for msg in messages:
+            clean_msg = {
+                "role": msg["role"],
+                "content": msg["content"]
+            }
+            clean_messages.append(clean_msg)
+        
         data = {
             "model": model,
-            "messages": messages,
+            "messages": clean_messages,
             "temperature": temperature,
             "max_tokens": max_tokens
         }
         
+        # Updated endpoint with correct path
         response = requests.post(
-            "https://endpoints.ai.cloud.ovh.net/v1/chat/completions",
+            "https://api.ai.cloud.ovh.net/v1/chat/completions",
             headers=headers,
             json=data
         )
@@ -962,21 +986,28 @@ def call_ovh_api(payload, api_key_override=None):
 def call_cerebras_api(payload, api_key_override=None):
     """Make a call to Cerebras API with error handling"""
     try:
-        # Use vanilla requests for this API
+        # Extract parameters from payload
         model = payload.get("model", "cerebras/llama-3.1-8b")
-        messages = payload.get("messages", [])
-        temperature = payload.get("temperature", 0.7)
-        max_tokens = payload.get("max_tokens", 1000)
-                
-        headers = {
-            "Content-Type": "application/json"
-        }
         
+        # Clean up messages - remove any unexpected properties
+        messages = []
+        for msg in payload.get("messages", []):
+            clean_msg = {
+                "role": msg["role"],
+                "content": msg["content"]
+            }
+            messages.append(clean_msg)
+            
         data = {
             "model": model,
             "messages": messages,
-            "temperature": temperature,
-            "max_tokens": max_tokens
+            "temperature": payload.get("temperature", 0.7),
+            "max_tokens": payload.get("max_tokens", 1000)
+        }
+        
+        headers = {
+            "Content-Type": "application/json",
+            "Authorization": f"Bearer {api_key_override or os.environ.get('CEREBRAS_API_KEY', '')}"
         }
         
         response = requests.post(