refresh

app.py

@@ -1,159 +1,270 @@
 import streamlit as st
 import os
-import asyncio
-from langchain_community.vectorstores import Chroma
-from langchain_community.embeddings import HuggingFaceBgeEmbeddings
+from langchain.vectorstores import Chroma
+from langchain.embeddings import HuggingFaceBgeEmbeddings
 from langchain_together import Together
 from langchain import hub
 from operator import itemgetter
+from langchain.schema.runnable import RunnableParallel
 from langchain.schema import format_document
-from langchain.prompts import ChatPromptTemplate, PromptTemplate
-from langchain_community.chat_message_histories import StreamlitChatMessageHistory
+from typing import List, Tuple
+from langchain.chains import LLMChain
+from langchain.chains import RetrievalQA
+from langchain.schema.output_parser import StrOutputParser
+from langchain.memory import StreamlitChatMessageHistory
 from langchain.memory import ConversationBufferMemory
-from langchain_core.runnables import RunnableLambda, RunnableParallel, RunnablePassthrough
+from langchain.chains import ConversationalRetrievalChain
+from langchain.memory import ConversationSummaryMemory
+from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder, PromptTemplate
+from langchain.schema.runnable import RunnableLambda, RunnablePassthrough
+
 
 # Load the embedding function
 model_name = "BAAI/bge-base-en"
-encode_kwargs = {'normalize_embeddings': True}
+encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity
+
 embedding_function = HuggingFaceBgeEmbeddings(
     model_name=model_name,
     encode_kwargs=encode_kwargs
 )
 
-# Initialize the LLMs
+# Load the ChromaDB vector store
+# persist_directory="./mrcpchromadb/"
+# vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function,collection_name="mrcppassmednotes")
+
+
+
+
+# Load the LLM
 llm = Together(
     model="mistralai/Mixtral-8x22B-Instruct-v0.1",
     temperature=0.2,
     top_k=12,
-    together_api_key=os.environ['pilotikval'],
-    max_tokens=200
+    together_api_key=os.environ['pilotikval']
 )
 
+# Load the summarizeLLM
 llmc = Together(
     model="mistralai/Mixtral-8x22B-Instruct-v0.1",
     temperature=0.2,
     top_k=3,
-    together_api_key=os.environ['pilotikval'],
-    max_tokens=200
+    together_api_key=os.environ['pilotikval']
 )
 
-# Memory setup
 msgs = StreamlitChatMessageHistory(key="langchain_messages")
 memory = ConversationBufferMemory(chat_memory=msgs)
 
-# Define the prompt templates
-CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(
-    """Given the following conversation and a follow-up question, rephrase the follow-up question to be a standalone question.
-    Chat History:
-    {chat_history}
-    Follow Up Input: {question}
-    Standalone question:"""
-)
 
-ANSWER_PROMPT = ChatPromptTemplate.from_template(
-    """You are helping a doctor. Answer based on the provided context:
-    {context}
-    Question: {question}"""
-)
 
-# Function to combine documents
-def _combine_documents(docs, document_prompt=PromptTemplate.from_template("{page_content}"), document_separator="\n\n"):
-    doc_strings = [format_document(doc, document_prompt) for doc in docs]
-    return document_separator.join(doc_strings)
+DEFAULT_DOCUMENT_PROMPT = PromptTemplate.from_template(template="{page_content}")
+
+def _combine_documents(
+        docs, document_prompt=DEFAULT_DOCUMENT_PROMPT, document_separator="\n\n"
+    ):
+        doc_strings = [format_document(doc, document_prompt) for doc in docs]
+        return document_separator.join(doc_strings)
+
+
 
-# Function to store chat history
 chistory = []
 
 def store_chat_history(role: str, content: str):
+    # Append the new message to the chat history
     chistory.append({"role": role, "content": content})
 
-# Define the chain using LCEL
-def create_conversational_qa_chain(retriever, condense_llm, answer_llm):
-    condense_question_chain = (
-        RunnableLambda(lambda x: {"chat_history": chistory, "question": x['question']})
-        | CONDENSE_QUESTION_PROMPT
-        | RunnableLambda(lambda x: {"standalone_question": x})
-    )
 
-    retrieval_chain = (
-        RunnableLambda(lambda x: {"standalone_question": x['standalone_question']})
-        | retriever
-        | RunnableLambda(lambda x: {"context": _combine_documents(x)})
-    )
-
-    answer_chain = ANSWER_PROMPT | answer_llm
+# Define the Streamlit app
+def app():
 
-    return RunnableParallel(
-        condense_question=condense_question_chain,
-        retrieve=retrieval_chain,
-        generate_answer=answer_chain
-    )
 
-# Asynchronous function to handle streaming responses
-async def stream_response(conversational_qa_chain, prompts2, chistory):
-    response_chunks = []
-    async for chunk in conversational_qa_chain.astream(
-        {
-            "question": prompts2,
-            "chat_history": chistory,
-        }
-    ):
-        response_chunks.append(chunk['generate_answer'])
-        st.write("".join(response_chunks))
-    return "".join(response_chunks)
 
-# Define the Streamlit app
-def app():
     with st.sidebar:
+
         st.title("dochatter")
+        # Create a dropdown selection box
         option = st.selectbox(
             'Which retriever would you like to use?',
             ('General Medicine', 'RespiratoryFishman', 'RespiratoryMurray', 'MedMRCP2', 'OldMedicine')
         )
+        # Depending on the selected option, choose the appropriate retriever
+        if option == 'RespiratoryFishman':
+            persist_directory="./respfishmandbcud/"
+            vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function,collection_name="fishmannotescud")
+            retriever = vectordb.as_retriever(search_kwargs={"k": 5})
+            retriever = retriever # replace with your actual retriever
         
-        # Define retrievers based on option
-        persist_directory = {
-            'General Medicine': "./oxfordmedbookdir/",
-            'RespiratoryFishman': "./respfishmandbcud/",
-            'RespiratoryMurray': "./respmurray/",
-            'MedMRCP2': "./medmrcp2store/",
-            'OldMedicine': "./mrcpchromadb/"
-        }.get(option, "./mrcpchromadb/")
-        
-        collection_name = {
-            'General Medicine': "oxfordmed",
-            'RespiratoryFishman': "fishmannotescud",
-            'RespiratoryMurray': "respmurraynotes",
-            'MedMRCP2': "medmrcp2notes",
-            'OldMedicine': "mrcppassmednotes"
-        }.get(option, "mrcppassmednotes")
-        
-        vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function, collection_name=collection_name)
-        retriever = vectordb.as_retriever(search_kwargs={"k": 5})
+        if option == 'RespiratoryMurray':
+            persist_directory="./respmurray/"
+            vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function,collection_name="respmurraynotes")
+            retriever = vectordb.as_retriever(search_kwargs={"k": 5})
+            retriever = retriever
+
+        if option == 'MedMRCP2':
+            persist_directory="./medmrcp2store/"
+            vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function,collection_name="medmrcp2notes")
+            retriever = vectordb.as_retriever(search_kwargs={"k": 5})
+            retriever = retriever
+
+        if option == 'General Medicine':
+            persist_directory="./oxfordmedbookdir/"
+            vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function,collection_name="oxfordmed")
+            retriever = vectordb.as_retriever(search_kwargs={"k": 7})
+            retriever = retriever
+            
+        else:
+            persist_directory="./mrcpchromadb/"
+            vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function,collection_name="mrcppassmednotes")
+            retriever = vectordb.as_retriever(search_kwargs={"k": 5})
+            retriever = retriever # replace with your actual retriever
+            retriever = retriever  # replace with your actual retriever
+
+        #template = """You are an AI chatbot having a conversation with a human. Use the following pieces of retrieved context to answer the question. If you don't know the answer, just say that you don't know. Use three sentences maximum and keep the answer concise.
+        #{context}
+        #{history}
+        #Human: {human_input}
+        #AI: """
+        #prompt = PromptTemplate(input_variables=["history", "question"], template=template)
+        #template = st.text_area("Template", value=template, height=180)
+        #prompt2 = ChatPromptTemplate.from_template(template)
 
-    if "messages" not in st.session_state:
+    
+    
+
+    # Session State
+    # Store LLM generated responses
+    if "messages" not in st.session_state.keys():
         st.session_state.messages = [{"role": "assistant", "content": "How may I help you?"}]
     
+
+    
+
+    
+
+
+    
+
+
+    ## Retry lets go
+
+    _template = """Given the following conversation and a follow up question, rephrase the follow up question to be a standalone question which contains the themes of the conversation. Do not write the question. Do not write the answer.
+    Chat History:
+    {chat_history}
+    Follow Up Input: {question}
+    Standalone question:"""
+    CONDENSE_QUESTION_PROMPT = PromptTemplate.from_template(_template)
+
+    template = """You are helping a doctor. Answer with what you know from the context provided. Please be as detailed and thorough. Answer the question based on the following context:
+    {context}
+    Question: {question}
+    """
+    ANSWER_PROMPT = ChatPromptTemplate.from_template(template)
+
+
+    _inputs = RunnableParallel(
+        standalone_question=RunnablePassthrough.assign(
+            chat_history=lambda x: chistory
+        )
+        | CONDENSE_QUESTION_PROMPT
+        | llmc
+        | StrOutputParser(),
+    )
+    _context = {
+        "context": itemgetter("standalone_question") | retriever | _combine_documents,
+        "question": lambda x: x["standalone_question"],
+    }
+    conversational_qa_chain = _inputs | _context | ANSWER_PROMPT | llm
+
+
+    
+
+    
+
+
+    
+    
+   
+    
     st.header("Ask Away!")
+    # Display the messages
     for message in st.session_state.messages:
         with st.chat_message(message["role"]):
             st.write(message["content"])
             store_chat_history(message["role"], message["content"])
+
+    # prompt = hub.pull("rlm/rag-prompt")
+    
+    
+    
     
     prompts2 = st.chat_input("Say something")
 
+    # Implement using different book sources, if statements
+    
+
+    
+
+    
+    
+
+    
+
+    
+
+    
+    
+    
     if prompts2:
         st.session_state.messages.append({"role": "user", "content": prompts2})
         with st.chat_message("user"):
             st.write(prompts2)
+            
+    
 
     if st.session_state.messages[-1]["role"] != "assistant":
-        conversational_qa_chain = create_conversational_qa_chain(retriever, llmc, llm)
         with st.chat_message("assistant"):
             with st.spinner("Thinking..."):
-                final_response = asyncio.run(stream_response(conversational_qa_chain, prompts2, chistory))
-        message = {"role": "assistant", "content": final_response}
+                response = conversational_qa_chain.invoke(
+                {
+                    "question": prompts2,
+                    "chat_history": chistory,
+                }
+            )
+                st.write(response)
+        message = {"role": "assistant", "content": response}
         st.session_state.messages.append(message)
 
+
+
+    
+        
+        
+        
+        
+
+
+   
+
+    # Create a button to submit the question
+    
+
+
+    
+
+        
+        
+
+        
+        
+        
+
+        
+        
+
+        
+
+    
+# Initialize history
+history = []
+
 if __name__ == '__main__':
-    app()
+    app()