app.py

import os

import streamlit as st

from langchain.chains import RetrievalQA
from langchain.document_loaders import PyPDFLoader

from langchain.vectorstores import Chroma
from langchain.document_loaders import TextLoader
from langchain.embeddings.openai import OpenAIEmbeddings

st.set_page_config(page_title="CoreMind AI", layout="wide")

st.header("CoreMind AI")

# ====================================================================================================

# SIDEBAR
st.sidebar.title("Options")

openai_key = st.sidebar.text_input("OpenAI API Key", type="password", key="openai_api_key")

os.environ["OPENAI_API_KEY"] = openai_key

# just go directly to the llm
from langchain.llms import OpenAI
# llm = OpenAI(temperature=0.9)

# st.write(llm("What is Berkshire Hathaway?"))


# ====================================================================================================

# chains do stuff in a prespecified order

# from langchain.prompts import PromptTemplate

# prompt = PromptTemplate(
#     input_variables=["product"],
#     template="Would Berkshire Hathaway typically invest in {product}?",
# )

# st.write(prompt.format(product="colorful socks"))

# from langchain.chains import LLMChain
# chain = LLMChain(llm=llm, prompt=prompt)

# st.write(chain.run("colorful socks"))


# ====================================================================================================

# agents decide what to do

# from langchain.agents import load_tools
# from langchain.agents import initialize_agent
# from langchain.agents import AgentType
from langchain.llms import OpenAI

# First, let's load the language model we're going to use to control the agent.
# llm = OpenAI(temperature=0)

# # Next, let's load some tools to use. Note that the `llm-math` tool uses an LLM, so we need to pass that in.
# tools = load_tools(["requests", "llm-math"], llm=llm)


# # Finally, let's initialize an agent with the tools, the language model, and the type of agent we want to use.
# agent = initialize_agent(tools, llm, agent=AgentType.ZERO_SHOT_REACT_DESCRIPTION, verbose=True)

# # Now let's test it out!
# st.write(agent.run("Is there a Q2 XXX? what was the 1987 relative perf?"))


# # ====================================================================================================


# from langchain.chat_models import ChatOpenAI

# this is more of a chat model, 

# whats' cool is you can batch messages
# batch_messages = [
#     [
#         SystemMessage(content="You are a helpful assistant that translates English to French."),
#         HumanMessage(content="I love programming.")
#     ],
#     [
#         SystemMessage(content="You are a helpful assistant that translates English to French."),
#         HumanMessage(content="I love artificial intelligence.")
#     ],
# ]
# result = chat.generate(batch_messages)
# result
# # -> LLMResult(generations=[[ChatGeneration(text="J'aime programmer.", generation_info=None, message=AIMessage(content="J'aime programmer.", additional_kwargs={}))], [ChatGeneration(text="J'aime l'intelligence artificielle.", generation_info=None, message=AIMessage(content="J'aime l'intelligence artificielle.", additional_kwargs={}))]], llm_output={'token_usage': {'prompt_tokens': 57, 'completion_tokens': 20, 'total_tokens': 77}})

# from langchain.embeddings.openai import OpenAIEmbeddings
# from langchain.vectorstores import Chroma
# from langchain.document_loaders import TextLoader

# loader = TextLoader("raw_data.txt")
# embeddings = OpenAIEmbeddings()
# docsearch = Chroma(persist_directory="data", embedding_function=embeddings)


# qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=docsearch.as_retriever())

# st.write(qa("What is Berkshire Hathaway?"))

# st.markdown("----")



if openai_key:
    def question_answer(user_text, qa_temperature):
        qa = RetrievalQA.from_chain_type(
            llm=OpenAI(temperature=qa_temperature, model_name="gpt-3.5-turbo"), 
            retriever=docsearch.as_retriever()
        )
        response = qa.run(user_text)
        return response
    
    
    loader = TextLoader("raw_data.txt")
    embeddings = OpenAIEmbeddings()
    docsearch = Chroma(persist_directory="data", embedding_function=embeddings)

    qa_tab, understanding_tab = st.tabs(["Database Understanding", "PDF Understanding"])

    with qa_tab:
        st.header("Question Answering")
        st.write("Dataset is Berkshire Hathaway's end of year reports for 1995, 1996 and 1997.")

        qa_query = st.text_area("Enter your query", key="qa_query", help="Got a question you think your docs can answer? Just ask!")
        
        if qa_query:
            response = question_answer(qa_query, 0.9)            
            st.write(response)



    with understanding_tab:
        llm = OpenAI(temperature=0.9)

        pdf_file = st.file_uploader("Upload a PDF", type=["pdf"], key="pdf_file")
        if pdf_file:
            with open("tmp_file.pdf", "wb") as f:
                f.write(pdf_file.getbuffer())

            loader = PyPDFLoader("tmp_file.pdf")
            pages = loader.load_and_split()

            from langchain.vectorstores import FAISS
            from langchain.embeddings.openai import OpenAIEmbeddings

            faiss_index = FAISS.from_documents(pages, OpenAIEmbeddings())
            # docs = faiss_index.similarity_search("How will the community be engaged?", k=2)
            # for doc in docs:
            #     st.write(str(doc.metadata["page"]) + ":", doc.page_content[:300])

            qa_prompt = st.text_area("Query your pdf", key="qa_prompt")
            # qa_button = st.button("Let's go!", disabled=not (openai_key and qa_prompt), key="qa_button", help="Make sure you have entered your OpenAI API key and a query.")

            if qa_prompt:

                qa = RetrievalQA.from_chain_type(llm=llm, chain_type="stuff", retriever=faiss_index.as_retriever())
                st.write(qa(qa_prompt))