updates

chat_dov.py

@@ -86,7 +86,6 @@ def create_chatbot(corpus, character_name, chatbot_type, retrieval_docs, summary
 ##  python -m streamlit run chat_dov.py -- --corpus data/tzamir.txt --character_name Dov --chatbot_type retrieval --retrieval_docs raw --interface streamlit
 
 def main():
-    os.environ["LANGCHAIN_HANDLER"] = "langchain"
 
     # parametros fixos para Dov Tzamir, arquivos ja processados , exceto indice que são em memoria
     st.title("Converse com o avatar do Dov Tzamir")
@@ -101,15 +100,10 @@ def main():
         "map_reduce",        #args.summary_type,
     )
 
-    openai_api_key = st.text_input(
-        label="Your OpenAI API KEY",
-        placeholder="Your OpenAI API KEY",
-        type="password",
-    )
     st.write(" ")
     st.write("Digite o seu diálogo aqui finalizando a linha com ENTER")
     st.write("Voce pode continuar o diálogo, apagando sua perguntanda anterior e  digitando aqui novamente")
-    #openai_api_key = os.environ["OPENAI_API_KEY"] 
+    openai_api_key = os.environ["OPENAI_API_KEY"] 
 
 
     app = Streamlit(chatbot=chatbot)

data_driven_characters/chatbots/retrieval.py

@@ -16,6 +16,8 @@ from data_driven_characters.memory import ConversationVectorStoreRetrieverMemory
 
 from langchain.embeddings.openai import OpenAIEmbeddings
 from langchain.vectorstores import FAISS
+import pickle
+import os.path
 
 
 class RetrievalChatBot:
@@ -34,36 +36,40 @@ class RetrievalChatBot:
         conv_memory = ConversationBufferMemory(
             memory_key=self.chat_history_key, input_key=self.input_key
         )
-        embeddings = OpenAIEmbeddings()
-        saved_db = FAISS.load_local('tzamir.ifass', embeddings)
-
-        """ DENTRO DO COMANDO SEGUINTE
+        #embeddings = OpenAIEmbeddings()
+        #saved_db = FAISS.load_local('tzamir.ifass', embeddings)
+        context_memory = ConversationVectorStoreRetrieverMemory(
             retriever=FAISS(
                 OpenAIEmbeddings().embed_query,
                 faiss.IndexFlatL2(1536),  # Dimensions of the OpenAIEmbeddings
                 InMemoryDocstore({}),
                 {},
             ).as_retriever(search_kwargs=dict(k=self.num_context_memories)),
-        """
-
-
-        context_memory = ConversationVectorStoreRetrieverMemory(
-            
-            
-            retriever=saved_db.as_retriever(search_kwargs=dict(k=self.num_context_memories)),
+            #retriever=saved_db.as_retriever(search_kwargs=dict(k=self.num_context_memories)),
             memory_key=self.context_key,
             output_prefix=character_definition.name,
             blacklist=[self.chat_history_key],
         )
 
-
-
-        # add the documents to the context memory
-        for i, summary in tqdm(enumerate(self.documents)):
-            context_memory.save_context(inputs={}, outputs={f"[{i}]": summary})
-
+        # add the documents to the context memory if not saved on disk
+        memory_path = 'output/tzamir/memory.pkl'
+        if not os.path.exists(memory_path):
+            print("gerando os indices")
+            for i, summary in tqdm(enumerate(self.documents)):
+                context_memory.save_context(inputs={}, outputs={f"[{i}]": summary})
+            # salvando no disco    
+            memory_pickle = open('output/tzamir/memory.pkl', 'wb') 
+            pickle.dump(context_memory, memory_pickle)
+        else:
+            print("carregando memoria do disco")
+            memory_pickle = open('output/tzamir/memory.pkl', 'rb') 
+            context_memory = pickle.load(memory_pickle)
         # Combined
         memory = CombinedMemory(memories=[conv_memory, context_memory])
+        #print("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$")
+        #print(memory)
+        #print("$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$$")
+    
         prompt = PromptTemplate.from_template(
             f"""Your name is {character_definition.name}.
 

output/tzamir/memory.pkl

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:0e35b2209f9a6ecc414572f31ea5f83a450ae259f524c53681ff7b12b1b9a80a
+size 719469