cgpt latest check

app.py

@@ -3,66 +3,28 @@ import os
 from langchain_community.vectorstores import Chroma
 from langchain_community.embeddings import HuggingFaceBgeEmbeddings
 from langchain_community.llms import Together
-from langchain import hub
-from operator import itemgetter
-from langchain.schema.runnable import RunnableParallel
+from langchain.prompts import ChatPromptTemplate, PromptTemplate
 from langchain.schema import format_document
-from typing import List, Tuple
-from langchain.chains import LLMChain
-from langchain.chains import RetrievalQA
-from langchain.schema.output_parser import StrOutputParser
-from langchain_community.chat_message_histories import StreamlitChatMessageHistory
+from typing import List
 from langchain.memory import ConversationBufferMemory
-from langchain.chains import ConversationalRetrievalChain
-from langchain.memory import ConversationSummaryMemory
-from langchain.prompts import ChatPromptTemplate, MessagesPlaceholder, PromptTemplate
-from langchain.schema.runnable import RunnableLambda, RunnablePassthrough
+from langchain.schema.runnable import RunnableParallel, RunnablePassthrough, StrOutputParser
+from langchain_core.chat_message_histories import StreamlitChatMessageHistory
 import time
 
 # Load the embedding function
 model_name = "BAAI/bge-base-en"
 encode_kwargs = {'normalize_embeddings': True} # set True to compute cosine similarity
-
-embedding_function = HuggingFaceBgeEmbeddings(
-    model_name=model_name,
-    encode_kwargs=encode_kwargs
-)
+embedding_function = HuggingFaceBgeEmbeddings(model_name=model_name, encode_kwargs=encode_kwargs)
 
 # Load the LLM
-llm = Together(
-    model="mistralai/Mixtral-8x22B-Instruct-v0.1",
-    temperature=0.2,
-    max_tokens=19096,
-    top_k=10,
-    together_api_key=os.environ['pilotikval']
-)
-
-# Load the summarizeLLM
-llmc = Together(
-    model="mistralai/Mixtral-8x7B-Instruct-v0.1",
-    temperature=0.2,
-    max_tokens=1024,
-    top_k=1,
-    together_api_key=os.environ['pilotikval']
-)
-
-# Load the reranking model
-reranker = Together(
-    model="mistralai/Mixtral-8x22B-Instruct-v0.1",
-    temperature=0.2,
-    max_tokens=512,
-    top_k=10,
-    together_api_key=os.environ['pilotikval']
-)
+llm = Together(model="mistralai/Mixtral-8x22B-Instruct-v0.1", temperature=0.2, max_tokens=19096, top_k=10, together_api_key=os.environ['pilotikval'], streaming=True)
 
 msgs = StreamlitChatMessageHistory(key="langchain_messages")
 memory = ConversationBufferMemory(chat_memory=msgs)
 
 DEFAULT_DOCUMENT_PROMPT = PromptTemplate.from_template(template="{page_content}")
 
-def _combine_documents(
-        docs, document_prompt=DEFAULT_DOCUMENT_PROMPT, document_separator="\n\n"
-    ):
+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)
 
@@ -76,7 +38,9 @@ def render_message_with_copy_button(role: str, content: str, key: str):
     html_code = f"""
     <div class="message" style="position: relative; padding-right: 40px;">
         <div class="message-content">{content}</div>
-        <button onclick="copyToClipboard('{key}')" style="position: absolute; right: 0; top: 0;">Copy</button>
+        <button onclick="copyToClipboard('{key}')" style="position: absolute; right: 0; top: 0; background-color: transparent; border: none; cursor: pointer;">
+            <img src="https://img.icons8.com/material-outlined/24/grey/copy.png" alt="Copy">
+        </button>
     </div>
     <textarea id="{key}" style="display:none;">{content}</textarea>
     <script>
@@ -92,16 +56,28 @@ def render_message_with_copy_button(role: str, content: str, key: str):
     """
     st.write(html_code, unsafe_allow_html=True)
 
-# Define the Streamlit app
+def get_streaming_response(user_query, chat_history):
+    template = """
+    You are a knowledgeable assistant. Provide a detailed and thorough answer to the question based on the following context:
+
+    Chat history: {chat_history}
+
+    User question: {user_question}
+    """
+    prompt = ChatPromptTemplate.from_template(template)
+    
+    inputs = {
+        "chat_history": chat_history,
+        "user_question": user_query
+    }
+
+    chain = prompt | llm | StrOutputParser()
+    return chain.stream(inputs)
+
 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
+        option = st.selectbox('Which retriever would you like to use?', ('General Medicine', 'RespiratoryFishman', 'RespiratoryMurray', 'MedMRCP2', 'OldMedicine'))
         if option == 'RespiratoryFishman':
             persist_directory = "./respfishmandbcud/"
             vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function, collection_name="fishmannotescud")
@@ -123,91 +99,30 @@ def app():
             vectordb = Chroma(persist_directory=persist_directory, embedding_function=embedding_function, collection_name="mrcppassmednotes")
             retriever = vectordb.as_retriever(search_kwargs={"k": 5})
 
-    # Session State
     if "messages" not in st.session_state.keys():
         st.session_state.messages = [{"role": "assistant", "content": "How may I help you?"}]
 
-    _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!")
     for i, message in enumerate(st.session_state.messages):
         with st.chat_message(message["role"]):
             render_message_with_copy_button(message["role"], message["content"], key=f"message-{i}")
             store_chat_history(message["role"], message["content"])
 
-    prompts2 = st.chat_input("Say something")
-
-    if prompts2:
-        st.session_state.messages.append({"role": "user", "content": prompts2})
+    user_query = st.chat_input("Say something")
+    if user_query:
+        st.session_state.messages.append({"role": "user", "content": user_query})
         with st.chat_message("user"):
-            st.write(prompts2)
+            st.write(user_query)
 
-    if st.session_state.messages[-1]["role"] != "assistant":
         with st.chat_message("assistant"):
             with st.spinner("Thinking..."):
-                for _ in range(3):  # Retry up to 3 times
-                    try:
-                        responses = generate_multiple_responses(
-                            conversational_qa_chain,
-                            {
-                                "question": prompts2,
-                                "chat_history": "\n".join([f"{msg['role']}: {msg['content']}" for msg in chistory])
-                            },
-                            num_responses=5
-                        )
-                        best_response = rerank_responses(reranker, responses)
-                        st.write(best_response)
-                        message = {"role": "assistant", "content": best_response}
-                        st.session_state.messages.append(message)
-                        break
-                    except Exception as e:
-                        st.error(f"An error occurred: {e}")
-                        time.sleep(2)  # Wait 2 seconds before retrying
-
-def generate_multiple_responses(chain, inputs, num_responses=5):
-    responses = []
-    for _ in range(num_responses):
-        response = chain.invoke(inputs)
-        responses.append(response)
-    return responses
-
-def rerank_responses(reranker, responses):
-    scores = []
-    for response in responses:
-        score = reranker.invoke(response)
-        scores.append(score)
-    best_response_idx = scores.index(max(scores))
-    return responses[best_response_idx]
-
-def stream_conversational_qa_chain(chain, inputs):
-    try:
-        response = chain.invoke(inputs)
-        for part in response:
-            yield part
-    except Exception as e:
-        raise e
+                chat_history = "\n".join([f"{msg['role']}: {msg['content']}" for msg in chistory])
+                response_generator = get_streaming_response(user_query, chat_history)
+                response_text = ""
+                for response_part in response_generator:
+                    response_text += response_part
+                    st.write(response_text)
+                st.session_state.messages.append({"role": "assistant", "content": response_text})
 
 if __name__ == '__main__':
     app()