Update app.py

app.py

@@ -11,41 +11,45 @@ from langchain_pinecone import PineconeVectorStore
 st.set_page_config('Opositor')
 st.header("Pregunta al trebep")
 
-#@st.cache_resource 
-def read():
+modelos_llm = [
+    'llama3-70b-8192',
+    'llama3-8b-8192',
+    'mixtral-8x7b-32768',
+    'gemma-7b-it'
+]
+modelo_llm = st.selectbox('Modelo de lenguaje', list(modelos_llm))
+
+@st.cache_resource 
+def setup(modelo_llm):
+    # Langsmith
+    os.environ["LANGCHAIN_TRACING_V2"] = "true"
+    os.environ["LANGCHAIN_API_KEY"] = "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"
+    os.environ["LANGCHAIN_PROJECT"] = "trebep"
+
     # CARGAMOS MODELO DE EMBEDDING
     model_name = 'intfloat/multilingual-e5-base'
     embedding = HuggingFaceEmbeddings(model_name=model_name)
+
+    # CARGAMOS LLM
+    os.environ["GROQ_API_KEY"] = "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"
+    llm = ChatGroq(model = modelo_llm)
     
-    return embedding
-
-# Langsmith
-os.environ["LANGCHAIN_TRACING_V2"] = "true"
-os.environ["LANGCHAIN_API_KEY"] = "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"
-os.environ["LANGCHAIN_PROJECT"] = "trebep"
-
-# CARGAMOS LLM
-os.environ["GROQ_API_KEY"] = "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"
-model = 'llama3-70b-8192'
-llm = ChatGroq(model = model)
-
-# CARGAMOS MODELO DE EMBEDDING
-embedding = read()
-
-# CARGAMOS EL VECTORSTORE DE PINECONE
-os.environ["PINECONE_API_KEY"] ='4bf0b4cf-4ced-4f70-8977-d60bb8ae405a'
-index_name = "boe-intfloat-multilingual-e5-base"
-namespace = "trebep"
-vectorstore = PineconeVectorStore(index_name=index_name, namespace=namespace, embedding=embedding)
-
-# CREAMOS EL RETRIEVAL
-qa = RetrievalQA.from_chain_type(
-    llm=llm,
-    chain_type="stuff",
-    retriever=vectorstore.as_retriever(),
-    return_source_documents=True,
-    #verbose=True
-)
+    # CARGAMOS EL VECTORSTORE DE PINECONE
+    os.environ["PINECONE_API_KEY"] ='4bf0b4cf-4ced-4f70-8977-d60bb8ae405a'
+    index_name = "boe-intfloat-multilingual-e5-base"
+    namespace = "trebep"
+    vectorstore = PineconeVectorStore(index_name=index_name, namespace=namespace, embedding=embedding)
+
+    # CREAMOS EL RETRIEVAL
+    qa = RetrievalQA.from_chain_type(
+        llm=llm,
+        chain_type="stuff",
+        retriever=vectorstore.as_retriever(),
+        return_source_documents=True,
+        #verbose=True
+    )
+
+    return qa
 
 # Función para mostrar logs
 def mostrar_logs(logs,hints):
@@ -68,6 +72,7 @@ user_question = st.text_input("¡A jugar! Haz una pregunta al trebep:")
 if user_question:
 
     # Inicializar entorno
+    qa = setup(modelo_llm)
 
     # Obtenemos la respuesta
     respuesta = qa.invoke(user_question)