Implement model integration strategy and selector for multiple AI models

model/ModelIntegrations.py

@@ -0,0 +1,37 @@
+from .ModelStrategy import ModelStrategy
+
+from langchain_community.chat_models import ChatOpenAI
+from langchain_mistralai.chat_models import ChatMistralAI
+from langchain_anthropic import ChatAnthropic
+from langchain_ollama import ChatOllama
+
+class MistralModel(ModelStrategy):
+    def get_model(self, model_name):
+        return ChatMistralAI(model=model_name)
+    
+
+class OpenAIModel(ModelStrategy):
+    def get_model(self, model_name):
+        return ChatOpenAI(model=model_name)
+
+
+class AnthropicModel(ModelStrategy):
+    def get_model(self, model_name):
+        return ChatAnthropic(model=model_name)
+
+
+class OllamaModel(ModelStrategy):
+    def get_model(self, model_name):
+        return ChatOllama(model=model_name)
+
+class ModelManager():
+    def __init__(self):
+        self.models = {
+            "mistral": MistralModel(),
+            "openai": OpenAIModel(),
+            "anthropic": AnthropicModel(),
+            "ollama": OllamaModel()
+        }
+
+    def get_model(self, provider, model_name):
+        return self.models[provider].get_model(model_name)

model/ModelStrategy.py

@@ -0,0 +1,6 @@
+from abc import ABC, abstractmethod
+
+class ModelStrategy(ABC):
+    @abstractmethod
+    def get_model(self, model_name):
+        pass

model/selector.py

@@ -0,0 +1,43 @@
+import streamlit as st
+from .ModelIntegrations import ModelManager
+    
+def ModelSelector():
+    # Dictionnaire des modèles par fournisseur
+    model_providers = {
+        "Mistral": {
+            "mistral-large-latest": "mistral.mistral-large-latest",
+            "open-mixtral-8x7b": "mistral.open-mixtral-8x7b",
+        },
+        "OpenAI": {
+            "gpt-4o": "openai.gpt-4o",
+        },
+        "Anthropic": {
+            "claude-3-5-sonnet-20240620": "anthropic.claude-3-5-sonnet-20240620",
+            "claude-3-opus-20240229": "anthropic.claude-3-opus-20240229",
+            "claude-3-sonnet-20240229": "anthropic.claude-3-sonnet-20240229",
+        },
+        # "Ollama": {
+        #     "llama3": "ollama.llama3"
+        # }
+    }
+
+    # Créer une liste avec les noms de modèle, groupés par fournisseur (fournisseur - modèle)
+    model_options = []
+    model_mapping = {}
+
+    for provider, models in model_providers.items():
+        for model_name, model_instance in models.items():
+            option_name = f"{provider} - {model_name}"
+            model_options.append(option_name)
+            model_mapping[option_name] = model_instance
+
+    # Sélection d'un modèle via un seul sélecteur
+    selected_model_option = st.selectbox("Choisissez votre modèle", options=model_options)
+    
+    # Afficher le modèle sélectionné
+    st.write(f"Current model: {model_mapping[selected_model_option]}")
+
+    if(st.session_state["assistant"]):
+        splitter = model_mapping[selected_model_option].split(".")
+        st.session_state["assistant"].setModel(ModelManager().get_model(splitter[0], splitter[1]))
+    

pages/chatbot.py

@@ -1,5 +1,6 @@
 import streamlit as st
 from streamlit_chat import message
+from model import selector
 
 def display_messages():
     for i, (msg, is_user) in enumerate(st.session_state["messages"]):
@@ -22,6 +23,8 @@ def process_input():
 def page():
     st.subheader("Posez vos questions")
 
+    selector.ModelSelector()
+
     if "assistant" not in st.session_state:
         st.text("Assistant non initialisé")
 

rag.py

@@ -19,7 +19,6 @@ from prompt_template import base_template
 # load .env in local dev
 load_dotenv()
 env_api_key = os.environ.get("MISTRAL_API_KEY")
-llm_model = "open-mixtral-8x7b"
 
 class Rag:
     document_vector_store = None
@@ -28,7 +27,7 @@ class Rag:
 
     def __init__(self, vectore_store=None):
         
-        self.model = ChatMistralAI(model=llm_model)
+        # self.model = ChatMistralAI(model=llm_model)
         self.embedding = MistralAIEmbeddings(model="mistral-embed", mistral_api_key=env_api_key)
 
         self.text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=100, length_function=len)
@@ -73,11 +72,9 @@ class Rag:
         )
 
     def ask(self, query: str, messages: list):
-        
+        print(self.model)
         self.chain = self.prompt | self.model | StrOutputParser()
         
-        print("messages ", messages)
-
         # Retrieve the context document
         if self.retriever is None:
             documentContext = ''

requirements.txt

@@ -17,3 +17,6 @@ langchain-openai
 langchain-community
 langchain-pinecone
 langchain_mistralai
+langchain_anthropic
+langchain_ollama
+pyyaml