feature : dynamic variables

app.py

@@ -1,15 +1,23 @@
 import streamlit as st
-import dotenv
 import os
 
+from dotenv import load_dotenv
 from rag import Rag
 from vectore_store.PineconeConnector import PineconeConnector
 from vectore_store.VectoreStoreManager import VectoreStoreManager
 
-GROUP_NAME = "Groupe 1"
 
-def main():
-        
+load_dotenv()
+
+GROUP_NAME = os.environ.get("APP_NAME")
+
+def init_app():
+    
+    # Read the environment variable and create a dictionary
+    variables = os.environ.get('VARIABLES')
+    keys = variables.split(',')
+    data_dict = {key: '' for key in keys}  # Initialize with empty values
+
     if len(st.session_state) == 0:
         # Define Vectore store strategy
         pinecone_connector = PineconeConnector()
@@ -17,6 +25,12 @@ def main():
 
         st.session_state["messages"] = []
         st.session_state["assistant"] = Rag(vectore_store=vs_manager)
+        st.session_state["data_dict"] = data_dict
+
+
+def main():
+
+    init_app()
 
     st.set_page_config(page_title=GROUP_NAME)
 

pages/chatbot.py

@@ -14,7 +14,7 @@ def process_input():
 
 
         with st.session_state["thinking_spinner"], st.spinner(f"Je réfléchis"):
-            agent_text = st.session_state["assistant"].ask(user_text, st.session_state["messages"] if "messages" in st.session_state else [])
+            agent_text = st.session_state["assistant"].ask(user_text, st.session_state["messages"] if "messages" in st.session_state else [], variables=st.session_state["data_dict"])
 
         st.session_state["messages"].append((user_text, True))
         st.session_state["messages"].append((agent_text, False))

pages/form.py

@@ -3,4 +3,8 @@ import streamlit as st
 def page():
     st.subheader("Définissez vos paramètres")
 
+    for key in st.session_state.data_dict.keys():
+        value = st.text_input(label=key, value=st.session_state.data_dict[key])
+        st.session_state.data_dict[key] = value  # Update the session state with user input
+
 page()

prompt_template.py

@@ -1,5 +1,9 @@
 base_template = '''
 
+Paramètre 1 : {param_1}
+Paramètre 2 : {param_2}
+Paramètre 3 : {param_3}
+
 Documents partagées : {commonContext}
 Document de référence : {documentContext}
 

rag.py

@@ -71,7 +71,7 @@ class Rag:
             },
         )
 
-    def ask(self, query: str, messages: list):
+    def ask(self, query: str, messages: list, variables: dict = {}):
         print(self.model)
         self.chain = self.prompt | self.model | StrOutputParser()
         
@@ -84,12 +84,22 @@ class Rag:
         # Retrieve the VectoreStore
         contextCommon = self.vector_store.retriever(query, self.embedding)
 
-        return self.chain.invoke({
+        # Dictionnaire de base avec les variables principales
+        chain_input = {
             "query": query,
             "documentContext": documentContext,
             "commonContext": contextCommon,
             "messages": messages
-        })
+        }
+
+        # Suppression des valeurs nulles (facultatif)
+        chain_input = {k: v for k, v in chain_input.items() if v is not None}
+
+        # Ajout dynamique d'autres variables dans **extra_vars
+        chain_input.update(variables)
+        print("chain_input", chain_input)
+        
+        return self.chain.invoke(chain_input)
 
     def clear(self):
         self.document_vector_store = None