app.py

import streamlit as st
from openai import OpenAI
import glob

from langchain_community.vectorstores import Chroma
from langchain_core.output_parsers import StrOutputParser
from langchain_core.runnables import RunnablePassthrough
from langchain_openai import ChatOpenAI, OpenAIEmbeddings
from langchain_text_splitters import RecursiveCharacterTextSplitter
from langchain.callbacks import get_openai_callback
from langchain_core.prompts import ChatPromptTemplate
from langchain_core.runnables import RunnableParallel

from langchain_community.document_loaders import UnstructuredFileLoader

OPENAI_API_KEY = st.secrets["OPENAI_API_KEY"]

# Get all the filenames from the docs folder
files = glob.glob("./docs/*.txt")

# Load files into readable documents
docs = []
for file in files:
    loader = UnstructuredFileLoader(file)
    docs.append(loader.load()[0])

# Config
with st.sidebar:
    st.write(f"Injected documents: \n\n {'\n'.join('\n'+file for file in files)}")
    model = st.selectbox("Model name", ["gpt-3.5-turbo"], disabled=True)
    temperature = st.number_input("Temperature", value=0.0, min_value=0.0, step=0.2, max_value=1.0, placeholder=0.0)
    k = st.number_input("Number of documents to include", value=1, min_value=1, step=1, placeholder=1)
    if st.toggle("Splitting", value=False):
        chunk_size = st.number_input("Chunk size", value=750, step=250, placeholder=750) # Defines the chunks in amount of tokens in which the files are split. Also defines the amount of tokens that are feeded into the context. 
        chunk_overlap = st.number_input("Chunk overlap", value=0, step=10, placeholder=0)
        text_splitter = RecursiveCharacterTextSplitter(chunk_size=chunk_size, chunk_overlap=chunk_overlap)
        splits = text_splitter.split_documents(docs)
        vectorstore = Chroma.from_documents(documents=splits, embedding=OpenAIEmbeddings())
    else:
        vectorstore = Chroma.from_documents(documents=docs, embedding=OpenAIEmbeddings())


prompt_template ="""
You are called "Volker". You are an assistant for question-answering tasks. 
You only answer questions about Long-Covid (use Post-Covid synonymously) and the Volker-App. 
If you don't know the answer, just say that you don't know. Say why you don't know the answer. 
Never answer questions about other diseases (e.g. Cancer-related fatigue, Multiple Sklerose).
Always answer in german language. Stay emphatic and positive.
When you use the word e.g "Arzt", "Ärzt", always write it as "Arzt".
Only use the following pieces of retrieved context to answer the question.
Question: {question} 
Context: {context} 
Answer:
""" # Source: hub.pull("rlm/rag-prompt")

# (1) Retriever
retriever = vectorstore.as_retriever(search_type="similarity_score_threshold", search_kwargs={"score_threshold": 0.3, "k": k})

# (2) Prompt
prompt = ChatPromptTemplate.from_template(prompt_template)

# (3) LLM
# Define the LLM we want to use. Default is "gpt-3.5-turbo" with temperature 0. 
# Temperature is a number between 0 and 1. With 0.8 it generates more random answers, with 0.2 it is more focused on the retrieved content. With temperature = 0 it uses log-probabilities depending on the content.

llm = ChatOpenAI(model_name=model, temperature=temperature)

def format_docs(docs):
    return "\n\n".join(doc.page_content for doc in docs)


# rag_chain = (
#     {"context": retriever | format_docs, "question": RunnablePassthrough()}
#     | prompt
#     | llm
#     | StrOutputParser()
# )



rag_chain_from_docs = (
    RunnablePassthrough.assign(context=(lambda x: format_docs(x["context"])))
    | prompt
    | llm
    | StrOutputParser()
)

rag_chain = RunnableParallel(
    {"context": retriever, "question": RunnablePassthrough()}
).assign(answer=rag_chain_from_docs)


st.title("🐔 Volker-Chat")

def click_button(prompt):
    st.session_state.clicked = True
    st.session_state['prompt'] = prompt


c = st.container()
c.write("Beispielfragen")
col1, col2, col3 = c.columns(3)
col1.button("Mehr zu 'Lernen'", on_click=click_button, args=["Was macht die Säule 'Lernen' aus?"])
col1.button("Was macht die Volker-App?", on_click=click_button, args=["Was macht die Volker-App?"])
col2.button("Mehr zu 'Tracken'", on_click=click_button, args=["Was macht die Säule 'Tracken' aus?"])
col2.button("Welche Krankenkassen erstatten die App?", on_click=click_button, args=["Welche Krankenkassen erstatten die App?"])
col3.button("Mehr zu 'Handeln'", on_click=click_button, args=["Was macht die Säule 'Handeln' aus?"])

if 'clicked' not in st.session_state:
    st.session_state.clicked = False

if "messages" not in st.session_state:
    st.session_state["messages"] = [{"role": "assistant", "content": "Ahoi! Ich bin Volker. Wie kann ich dir helfen?"}]

for msg in st.session_state.messages:
    st.chat_message(msg["role"]).write(msg["content"])

if st.session_state.clicked:
    prompt = st.session_state['prompt']
    st.chat_message("user").write(prompt)
    with get_openai_callback() as cb:
        response = rag_chain.invoke(prompt)
        st.chat_message("assistant").write(response['answer'])
        with st.expander("Kontext ansehen"):
            st.write(response["context"])
        with st.sidebar:
            sidebar_c = st.container()
            sidebar_c.success(cb)


if prompt := st.chat_input():
    st.chat_message("user").write(prompt)
    with get_openai_callback() as cb:
        response = rag_chain.invoke(prompt)
        st.chat_message("assistant").write(response['answer'])
        with st.expander("Kontext ansehen"):
            st.write(response["context"])
        with st.sidebar:
            sidebar_c = st.container()
            sidebar_c.success(cb)


# cleanup
st.session_state.clicked = False
vectorstore.delete_collection()