add gemini routes

multi_llm_chatbot_backend/app/api/routes.py

@@ -1,6 +1,8 @@
+import os
 from fastapi import APIRouter, Body, HTTPException
 import httpx
 from app.llm.llm_client import LLMClient
+from app.llm.gemini_client import GeminiClient
 from app.models.persona import Persona
 from app.core.orchestrator import ChatOrchestrator
 from app.core.seamless_orchestrator import SeamlessOrchestrator
@@ -9,7 +11,28 @@ from typing import Optional, List
 
 router = APIRouter()
 
-# Improved LLM client with better short response handling
+# Provider management
+current_provider = "gemini"
+available_providers = ["ollama", "gemini"]
+
+def create_llm_client(provider: str = None) -> LLMClient:
+    """Create LLM client based on provider"""
+    if provider is None:
+        provider = current_provider
+        
+    if provider == "gemini":
+        try:
+            return GeminiClient(model_name="gemini-1.5-flash")
+        except ValueError as e:
+            # Fallback to Ollama if Gemini API key is not available
+            print(f"Gemini API key not found, falling back to Ollama: {e}")
+            return ShortResponseOllamaClient(model_name="llama3.2:1b")
+    elif provider == "ollama":
+        return ShortResponseOllamaClient(model_name="llama3.2:1b")
+    else:
+        raise ValueError(f"Unknown provider: {provider}")
+
+# Improved LLM client with better short response handling for Ollama
 class ShortResponseOllamaClient(LLMClient):
     def __init__(self, model_name: str = "llama3.2:1b"):
         self.model_name = model_name
@@ -44,7 +67,7 @@ class ShortResponseOllamaClient(LLMClient):
                 "temperature": 0.7,
                 "top_p": 0.9,
                 "top_k": 40,
-                "num_predict": 200,  # Increased from 150
+                "num_predict": 200,
                 "repeat_penalty": 1.1,
             }
         }
@@ -70,7 +93,8 @@ class ShortResponseOllamaClient(LLMClient):
         except Exception as e:
             return f"I apologize, but I'm having trouble generating a response right now. Please try again."
 
-llm = ShortResponseOllamaClient(model_name="llama3.2:1b")
+# Initialize with default provider
+llm = create_llm_client()
 chat_orchestrator = ChatOrchestrator()
 seamless_orchestrator = SeamlessOrchestrator(llm=llm)
 
@@ -83,7 +107,6 @@ class GlobalSessionContext:
         self.full_log.append({"role": role, "content": content})
 
     def filter_by_persona(self, persona_id: str):
-        # Return full context but mark the persona for better understanding
         return self.full_log
 
     def clear(self):
@@ -91,28 +114,31 @@ class GlobalSessionContext:
 
 session_context = GlobalSessionContext()
 
-# Improved personas with natural, conversational prompts
-DEFAULT_PERSONAS = [
-    Persona(
-        id="methodist",
-        name="Methodist Advisor",
-        system_prompt="You are Dr. Methodist, a structured PhD advisor. Give brief, organized advice in 2-3 clear sentences. Focus on systematic approaches and planning.",
-        llm=llm
-    ),
-    Persona(
-        id="theorist",
-        name="Theorist Advisor", 
-        system_prompt="You are Dr. Theorist, a philosophical PhD advisor. Give brief, thoughtful insights in 2-3 sentences. Focus on concepts, frameworks, and deeper understanding.",
-        llm=llm
-    ),
-    Persona(
-        id="pragmatist",
-        name="Pragmatist Advisor",
-        system_prompt="You are Dr. Pragmatist, a practical PhD advisor. Give brief, actionable advice in 2-3 sentences. Focus on concrete steps and real-world solutions.",
-        llm=llm
-    )
-]
-
+def create_default_personas(llm_client: LLMClient):
+    """Create default personas with given LLM client"""
+    return [
+        Persona(
+            id="methodist",
+            name="Methodist Advisor",
+            system_prompt="You are Dr. Methodist, a structured PhD advisor. Give brief, organized advice in 2-3 clear sentences. Focus on systematic approaches and planning.",
+            llm=llm_client
+        ),
+        Persona(
+            id="theorist",
+            name="Theorist Advisor", 
+            system_prompt="You are Dr. Theorist, a philosophical PhD advisor. Give brief, thoughtful insights in 2-3 sentences. Focus on concepts, frameworks, and deeper understanding.",
+            llm=llm_client
+        ),
+        Persona(
+            id="pragmatist",
+            name="Pragmatist Advisor",
+            system_prompt="You are Dr. Pragmatist, a practical PhD advisor. Give brief, actionable advice in 2-3 sentences. Focus on concrete steps and real-world solutions.",
+            llm=llm_client
+        )
+    ]
+
+# Initialize personas
+DEFAULT_PERSONAS = create_default_personas(llm)
 for persona in DEFAULT_PERSONAS:
     chat_orchestrator.register_persona(persona)
 
@@ -134,6 +160,63 @@ class ReplyToAdvisor(BaseModel):
     advisor_id: str
     original_message_id: Optional[str] = None
 
+class ProviderSwitch(BaseModel):
+    provider: str
+
+# Provider management endpoints
+@router.get("/current-provider")
+async def get_current_provider():
+    return {
+        "current_provider": current_provider,
+        "available_providers": available_providers,
+        "model_info": {
+            "name": llm.model_name if hasattr(llm, 'model_name') else "gemini-1.5-flash",
+            "provider": current_provider
+        }
+    }
+
+@router.post("/switch-provider")
+async def switch_provider(provider_data: ProviderSwitch):
+    global current_provider, llm
+    
+    if provider_data.provider not in available_providers:
+        raise HTTPException(
+            status_code=400, 
+            detail=f"Unknown provider: {provider_data.provider}. Available: {available_providers}"
+        )
+    
+    try:
+        # Update current provider
+        current_provider = provider_data.provider
+        
+        # Create new LLM client
+        new_llm = create_llm_client(current_provider)
+        llm = new_llm
+        
+        # Update all personas with new LLM
+        new_personas = create_default_personas(new_llm)
+        chat_orchestrator.personas.clear()
+        for persona in new_personas:
+            chat_orchestrator.register_persona(persona)
+        
+        # Update seamless orchestrator
+        seamless_orchestrator.llm = new_llm
+        
+        return {
+            "message": f"Successfully switched to {current_provider}",
+            "current_provider": current_provider,
+            "model_info": {
+                "name": new_llm.model_name if hasattr(new_llm, 'model_name') else "gemini-1.5-flash",
+                "provider": current_provider
+            }
+        }
+        
+    except Exception as e:
+        raise HTTPException(
+            status_code=500,
+            detail=f"Failed to switch to {provider_data.provider}: {str(e)}"
+        )
+
 # Sequential advisor responses endpoint
 @router.post("/chat-sequential")
 async def chat_sequential(message: ChatMessage):
@@ -184,7 +267,7 @@ async def chat_sequential(message: ChatMessage):
                     responses.append({
                         "persona": chat_orchestrator.personas[persona_id].name,
                         "persona_id": persona_id,
-                        "response": "I'm having trouble generating a response right now. Please try asking again.",
+                        "response": "I'm having trouble generating a response right now. Please try again.",
                         "order": i
                     })
 
@@ -204,6 +287,40 @@ async def chat_sequential(message: ChatMessage):
             }]
         }
 
+# Main chat endpoint (keep for compatibility)
+@router.post("/chat")
+async def chat_with_orchestrator(message: ChatMessage):
+    """Redirect to sequential endpoint for better UX"""
+    return await chat_sequential(message)
+
+# Individual advisor endpoint with context
+@router.post("/chat/{persona_id}")
+async def chat_with_specific_advisor(persona_id: str, input: UserInput):
+    """Chat with a specific advisor"""
+    try:
+        if persona_id not in chat_orchestrator.personas:
+            raise HTTPException(status_code=404, detail=f"Persona '{persona_id}' not found")
+
+        session_context.append("user", input.user_input)
+        persona = chat_orchestrator.personas[persona_id]
+        context = session_context.full_log.copy()
+        reply = await persona.respond(context)
+        session_context.append(persona_id, reply)
+
+        return {
+            "persona": persona.name,
+            "persona_id": persona_id,
+            "response": reply
+        }
+    except HTTPException:
+        raise
+    except Exception as e:
+        print(f"Error in chat_with_specific_advisor: {e}")
+        return {
+            "persona": "System",
+            "response": "I'm having trouble generating a response right now. Please try again."
+        }
+
 # Reply to specific advisor endpoint
 @router.post("/reply-to-advisor")
 async def reply_to_advisor(reply: ReplyToAdvisor):
@@ -242,40 +359,6 @@ async def reply_to_advisor(reply: ReplyToAdvisor):
             "response": "I'm having trouble generating a reply right now. Please try again."
         }
 
-# Main chat endpoint (keep for compatibility)
-@router.post("/chat")
-async def chat_with_orchestrator(message: ChatMessage):
-    """Redirect to sequential endpoint for better UX"""
-    return await chat_sequential(message)
-
-# Individual advisor endpoint with context
-@router.post("/chat/{persona_id}")
-async def chat_with_specific_advisor(persona_id: str, input: UserInput):
-    """Chat with a specific advisor"""
-    try:
-        if persona_id not in chat_orchestrator.personas:
-            raise HTTPException(status_code=404, detail=f"Persona '{persona_id}' not found")
-
-        session_context.append("user", input.user_input)
-        persona = chat_orchestrator.personas[persona_id]
-        context = session_context.full_log.copy()
-        reply = await persona.respond(context)
-        session_context.append(persona_id, reply)
-
-        return {
-            "persona": persona.name,
-            "persona_id": persona_id,
-            "response": reply
-        }
-    except HTTPException:
-        raise
-    except Exception as e:
-        print(f"Error in chat_with_specific_advisor: {e}")
-        return {
-            "persona": "System",
-            "response": "I'm having trouble generating a response right now. Please try again."
-        }
-
 # Reset session
 @router.post("/reset-session")
 async def reset_session():
@@ -292,26 +375,23 @@ async def reset_session():
 def get_context():
     return session_context.full_log
 
-# Model switching for development
+# Legacy model switching endpoint (now redirects to provider switching)
 @router.post("/switch-model")
 async def switch_model(model_name: str = Body(...)):
-    global llm
-    try:
-        llm = ShortResponseOllamaClient(model_name=model_name)
-        
-        # Update all personas with new LLM
-        for persona_id in chat_orchestrator.personas:
-            chat_orchestrator.personas[persona_id].llm = llm
-        
-        seamless_orchestrator.llm = llm
-        
-        return {"message": f"Switched to model: {model_name}"}
-    except Exception as e:
-        return {"error": f"Failed to switch model: {str(e)}"}
+    # For backward compatibility, try to map model names to providers
+    if "gemini" in model_name.lower():
+        return await switch_provider(ProviderSwitch(provider="gemini"))
+    else:
+        return await switch_provider(ProviderSwitch(provider="ollama"))
 
 @router.get("/current-model")
 async def get_current_model():
-    return {"model": llm.model_name}
+    # For backward compatibility
+    model_name = llm.model_name if hasattr(llm, 'model_name') else "gemini-1.5-flash"
+    return {
+        "model": model_name,
+        "provider": current_provider
+    }
 
 # Debug endpoint
 @router.get("/debug/personas")
@@ -323,5 +403,6 @@ async def debug_personas():
                 "prompt": persona.system_prompt[:100] + "..."
             } for pid, persona in chat_orchestrator.personas.items()
         },
-        "context_length": len(session_context.full_log)
+        "context_length": len(session_context.full_log),
+        "current_provider": current_provider
     }