streamlit_interface.py

import streamlit as st
from load_into_docs import git_repo_reader, clone_git_repo, load_git_from_disk, chunk_data_based_on_markdown

from process_gpt import ChatGPT
from create_vectorstore import CreateVectorStore

class StreamlitInterface:
    def __init__(self):
        self.repo_owner = None
        self.repo_name = None
        
    def start_interface(self):
        st.title("💻🦾 Load a GitHub repository")
        git_url = st.text_input("Enter the GitHub URL of the repository: ")
        if git_url:
            # Parse the git URL
            self.parse_git_url(git_url=git_url)

            # Load the repository from GitHub
            path_of_repo, branch = clone_git_repo(git_url=git_url)
            documents = load_git_from_disk(path_of_repo=path_of_repo, branch=branch)

            texts = chunk_data_based_on_markdown(documents)
            st.write(f"Number of documents: {len(documents)}, Number of texts: {len(texts)}")

            # create vectorstore
            st.write("Creating index...")
            create_vec_store = CreateVectorStore()
            docsearch = create_vec_store.create_chroma_vectorstore(texts=texts)

            #embeddings = create_embeddings()
            #docsearch = create_chroma_vectorstore(texts=texts, embeddings=embeddings)
            st.write("Index created.")

            # instantiate retriever
            chat_gpt = ChatGPT()
            chat_gpt.create_self_querying_retriever(
                docsearch, 
                create_vec_store.return_document_content_description(), 
                create_vec_store.return_metadata_field_info()
            )

            qa = chat_gpt.gpt_conversational_retrieval_chain()
            
            # Run gpt chatbot
            query = st.text_input("Enter a question: ")
            if query:
                res = chat_gpt.gpt_question_remember_history(qa, query)
                st.write(res["answer"])
                st.write(res["chat_history"])

                # gpt_answer = chat_gpt.gpt_answer(relevant_docs, query)
                # st.write(gpt_answer)
                st.write("Relevant documents:")
                for doc in res["source_documents"]:
                    with st.expander(doc.metadata["file_name"]):
                        st.info(doc.page_content)

    def parse_git_url(self, git_url):
        """Parse a git URL into the owner and repo name"""
        (self.repo_owner, self.repo_name) = git_url.split("/")[-2:]        

    def load_repo_from_github(self):
        """Load a repository from GitHub"""
        st.write(f"Loading the repository {self.repo_name}...")
        documents= git_repo_reader(self.repo_owner, self.repo_name)
        st.write("Repository loaded.")
        print(f'Number of documents: {len(documents)}')
        return documents