Update app.py

app.py

@@ -3,8 +3,9 @@ import os
 from dotenv import load_dotenv
 from PyPDF2 import PdfReader
 from langchain.text_splitter import CharacterTextSplitter
-from langchain_community.embeddings import HuggingFaceInstructEmbeddings
+from langchain_community.embeddings import HuggingFaceInstructEmbeddings,HuggingFaceEmbeddings,CohereEmbeddings
 from langchain_openai import OpenAIEmbeddings,ChatOpenAI
+from langchain_community.chat_models import ChatCohere
 from langchain_community.vectorstores import FAISS
 from langchain.memory import ConversationBufferMemory
 from langchain.chains import ConversationalRetrievalChain
@@ -21,7 +22,6 @@ import langchain
 langchain.verbose = False
 
 
-
 def get_pdf_text(pdf_docs):
     text = ""
     for pdf in pdf_docs:
@@ -41,35 +41,50 @@ def get_text_chunks(text):
     return chunks
 
 def get_vectorstore(text_chunks,selected_embedding):
+    print('Selected Embedding: ' + selected_embedding)
     if selected_embedding == 'OpenAI':
-        print('OpenAI embedding')
         embeddings = OpenAIEmbeddings()
     elif selected_embedding == 'Instructor-xl':
-        print('Instructor-xl embedding')
+        
         embeddings = HuggingFaceInstructEmbeddings(model_name="hkunlp/instructor-xl")
+    elif selected_embedding == 'Cohere-multilingual-v3.0':
+        
+        embeddings = CohereEmbeddings(
+            model="embed-multilingual-v3.0",
+            cohere_api_key=os.environ.get("COHERE_API_KEY")
+        )
 
     vectorstore = FAISS.from_texts(texts=text_chunks, embedding=embeddings)
     vectorstore.save_local("faiss_index")
     return vectorstore
 
 def load_vectorstore(text_chunks,selected_embedding):
+    print('Selected Embedding: ' + selected_embedding)
     if selected_embedding == 'OpenAI':
-        print('OpenAI embedding')
         embeddings = OpenAIEmbeddings()
     elif selected_embedding == 'Instructor-xl':
-        print('Instructor-xl embedding')
         embeddings = HuggingFaceInstructEmbeddings(model_name="hkunlp/instructor-xl")
         vectorstore = FAISS.load_local("faiss_index", embeddings)
+    elif selected_embedding == 'Cohere-multilingual-v3.0':
+   
+        embeddings = CohereEmbeddings(
+            model="embed-multilingual-v3.0",
+            cohere_api_key=os.environ.get("COHERE_API_KEY")
+        )
+    
+    vectorstore = FAISS.load_local("faiss_index", embeddings)
     return vectorstore
 
-def get_conversation_chain(vectorstore,selected_llm):
-    if selected_llm == 'OpenAI':
-        print('OpenAi LLM')
-        llm = ChatOpenAI()
-
+def get_conversation_chain(vectorstore,selected_llm,selected_temperature):
+    print('Seleted LLM: ' + selected_llm)
+    print('Selected Temperature: ' + str(selected_temperature))
     
+    if selected_llm == 'OpenAI':
+        #openai_model = "gpt-4-turbo-preview"
+        openai_model = "gpt-3.5-turbo"
+        llm = ChatOpenAI(model=openai_model,temperature=selected_temperature)
     elif selected_llm == 'Llama2':
-        print('Llama2 LLM')
+        
         model_id = 'meta-llama/Llama-2-7b-chat-hf'
         hf_auth = os.environ.get("HUGGINGFACEHUB_API_TOKEN") 
         
@@ -96,7 +111,7 @@ def get_conversation_chain(vectorstore,selected_llm):
                 config=model_config,
                 quantization_config=bnb_config,
                 device_map='auto',
-                token=hf_auth
+                token=os.environ.get("HUGGINGFACEHUB_API_TOKEN")
             )
         else:
             model = transformers.AutoModelForCausalLM.from_pretrained(
@@ -104,7 +119,7 @@ def get_conversation_chain(vectorstore,selected_llm):
                 trust_remote_code=True,
                 config=model_config,
                 device_map='auto',
-                token=hf_auth
+                token=os.environ.get("HUGGINGFACEHUB_API_TOKEN")
             )
 
         # enable evaluation mode to allow model inference
@@ -113,7 +128,7 @@ def get_conversation_chain(vectorstore,selected_llm):
 
         tokenizer = transformers.AutoTokenizer.from_pretrained(
             model_id,
-            token=hf_auth
+            token=os.environ.get("HUGGINGFACEHUB_API_TOKEN")
         )
 
         pipeline = transformers.pipeline(        
@@ -122,17 +137,20 @@ def get_conversation_chain(vectorstore,selected_llm):
             tokenizer=tokenizer,
             return_full_text=True,  # langchain expects the full text
             task='text-generation',
-            temperature=0.1,  # 'randomness' of outputs, 0.0 is the min and 1.0 the max
+            temperature=selected_temperature,  # 'randomness' of outputs, 0.0 is the min and 1.0 the max
             max_new_tokens=512,  # max number of tokens to generate in the output
             repetition_penalty=1.1  # without this output begins repeating
         )
 
         llm = HuggingFacePipeline(pipeline=pipeline)
 
+        
+
     # Generic LLM    
     memory = ConversationBufferMemory(
     memory_key='chat_history', return_messages=True)
-        
+
+           
     conversation_chain = ConversationalRetrievalChain.from_llm(
         llm=llm,
         retriever=vectorstore.as_retriever(),
@@ -178,8 +196,9 @@ def main():
         pdf_docs = st.file_uploader(
             "Upload your new PDFs here and click on 'Process' or load the last upload by clicking on 'Load'", accept_multiple_files=True)
 
-        selected_embedding = st.radio("Which Embedding?",["OpenAI", "Instructor-xl"])
+        selected_embedding = st.radio("Which Embedding?",["Cohere-multilingual-v3.0","OpenAI", "Instructor-xl"])
         selected_llm = st.radio("Which LLM?",["OpenAI", "Llama2"])
+        selected_temperature = st.slider('Temperature?', 0.0, 1.0, 0.1)
 
         if st.button("Process"):
             with st.spinner("Processing"):
@@ -194,7 +213,7 @@ def main():
 
                 # create conversation chain
                 st.session_state.conversation = get_conversation_chain(
-                    vectorstore,selected_llm)
+                    vectorstore,selected_llm,selected_temperature)
 
         if st.button("Load"):
             with st.spinner("Processing"):
@@ -204,7 +223,7 @@ def main():
 
                 # create conversation chain
                 st.session_state.conversation = get_conversation_chain(
-                    vectorstore,selected_llm)
+                    vectorstore,selected_llm,selected_temperature)
     
     if st.session_state.conversation:
         st.header("VerAi :books:")
@@ -216,4 +235,3 @@ def main():
 
 if __name__ == '__main__':
     main()
-