app.py

import streamlit as st
import torch
from langchain_text_splitters import Language, RecursiveCharacterTextSplitter
from langchain_community.embeddings import HuggingFaceEmbeddings
from langchain_community.vectorstores import FAISS
from langchain_community.llms.huggingface_pipeline import HuggingFacePipeline
from langchain.memory import ConversationBufferMemory
from langchain.chains import ConversationalRetrievalChain
from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline



# gpt_model = 'gpt-4-1106-preview'
# embedding_model = 'text-embedding-3-small'
default_model_id = "bigcode/starcoder2-3b"
#default_model_id = "tiiuae/falcon-7b-instruct"

def init():
    if "conversation" not in st.session_state:
        st.session_state.conversation = None
    if "chat_history" not in st.session_state:
        st.session_state.chat_history = []        

def init_llm_pipeline(model_id):
    if "llm" not in st.session_state:     

        model = AutoModelForCausalLM.from_pretrained(
            model_id,
            device_map="auto"
        )      
        tokenizer = AutoTokenizer.from_pretrained(model_id)
        tokenizer.add_eos_token = True
        tokenizer.pad_token_id = 0
        tokenizer.padding_side = "left"

        text_generation_pipeline = pipeline(
        model=model,
        tokenizer=tokenizer,
        task="text-generation",
        max_new_tokens=1024
        )
        st.session_state.llm = text_generation_pipeline         

def get_retriever(files):
    documents = [doc.getvalue().decode("utf-8") for doc in files]
    python_splitter = RecursiveCharacterTextSplitter.from_language(
        language=Language.PYTHON, chunk_size=2000, chunk_overlap=200
    )

    texts = python_splitter.create_documents(documents)

    embeddings = HuggingFaceEmbeddings(model_name="all-MiniLM-L6-v2")

    db = FAISS.from_documents(texts, embeddings)
    retriever = db.as_retriever(
        search_type="mmr",  # Also test "similarity"
        search_kwargs={"k": 8},
    )
    return retriever
    
def get_conversation(retriever):
    #memory = ConversationBufferMemory(memory_key='chat_history', return_messages=True)

    conversation_chain = ConversationalRetrievalChain.from_llm(
        llm=st.session_state.llm,
        retriever=retriever   
    )
    return conversation_chain

def getprompt(user_input):
    prompt = f"You are a helpful assistant. Please answer the user question. USER: {user_input} ASSISTANT:"
    return prompt

def handle_user_input(question):
    st.session_state.chat_history += {"role":"user","content":question}
    response = st.session_state.llm(getprompt(question))
    st.session_state.chat_history += {"role":"assistant","content":response}
    for i, message in enumerate(st.session_state.chat_history):
        if i % 2 == 0:
            with st.chat_message("user"):
                st.write(message.content)
        else:
            with st.chat_message("assistant"):
                st.write(message.content)

def main():
    init()

    st.set_page_config(page_title="Coding-Assistent", page_icon=":books:")

    st.header(":books: Coding-Assistent ")
    user_input = st.chat_input("Stellen Sie Ihre Frage hier")
    if user_input:
        with st.spinner("Führe Anfrage aus ..."):        
            handle_user_input(user_input)


    with st.sidebar:
        st.subheader("Model selector")
        model_id = st.text_input("Modelname on HuggingFace", default_model_id) 
        st.subheader("Code Upload")
        upload_docs=st.file_uploader("Dokumente hier hochladen", accept_multiple_files=True)
        if st.button("Hochladen"):
            with st.spinner("Analysiere Dokumente ..."):
                init_llm_pipeline(model_id)
                #retriever = get_retriever(upload_docs)
                #st.session_state.conversation = get_conversation(retriever) 


if __name__ == "__main__":
    main()