pratikshahp/openai-github-chat

1import streamlit as st2import os3from github import Github4from langchain_community.vectorstores import Chroma5from langchain_community.embeddings import HuggingFaceEmbeddings6from langchain_text_splitters import RecursiveCharacterTextSplitter7from openai import OpenAI8from dotenv import load_dotenv9 10# Load environment variables11load_dotenv()12openai_api_key = os.getenv("OPENAI_API_KEY")13 14# Function to fetch repository data from GitHub15def fetch_github_repo_data(repo_name, github_token):16    """Fetch all text content from a GitHub repository."""17    try:18        g = Github(github_token)19        repo = g.get_repo(repo_name)20        contents = repo.get_contents("")21        repo_data = ""22 23        while contents:24            file_content = contents.pop(0)25            if file_content.type == "dir":26                contents.extend(repo.get_contents(file_content.path))27            else:28                try:29                    file_data = repo.get_contents(file_content.path).decoded_content30                    text = file_data.decode("utf-8")31                    repo_data += f"\n\nFile: {file_content.path}\n{text}"32                except UnicodeDecodeError:33                    # Skip non-text files34                    continue35 36        return repo_data37    except Exception as e:38        st.error(f"Error fetching GitHub repository data: {e}")39        return None40 41# Function to generate a response using OpenAI42def generate_response(context, question):43    """Generate a response using OpenAI."""44    try:45        from openai import OpenAI46 47        client = OpenAI(api_key=openai_api_key)48        messages = [49            {"role": "system", "content": "You are an assistant that answers questions based on repository content."},50            {"role": "user", "content": f"Context: {context}\n\nQuestion: {question}\n\nAnswer:"}51        ]52        response = client.chat.completions.create(53            model="gpt-4o-mini",54            messages=messages,55            max_tokens=150,56        )57        return response.choices[0].message.content.strip()58    except Exception as e:59        st.error(f"Error generating response: {e}")60        return None61 62# Function to perform RAG using OpenAI and Chroma63def perform_rag(repo_data, question):64    """Perform retrieval-augmented generation using ChromaDB and OpenAI."""65    try:66        if not repo_data:67            st.warning("Repository data is empty.")68            return None69 70        # Create embeddings71        embeddings = HuggingFaceEmbeddings()72 73        # Split text into chunks74        text_splitter = RecursiveCharacterTextSplitter(75            chunk_size=1000, chunk_overlap=20, length_function=len76        )77        chunks = text_splitter.create_documents([repo_data])78 79        # Store chunks in ChromaDB80        persist_directory = "github_repo_embeddings"81        vectordb = Chroma.from_documents(82            documents=chunks, embedding=embeddings, persist_directory=persist_directory83        )84        vectordb.persist()85 86        # Load persisted Chroma database87        vectordb = Chroma(88            persist_directory=persist_directory, embedding_function=embeddings89        )90 91        # Perform retrieval using Chroma92        docs = vectordb.similarity_search(question)93        if not docs:94            st.warning("No relevant documents found.")95            return None96 97        context = docs[0].page_content98        return generate_response(context, question)99 100    except Exception as e:101        st.error(f"Error performing RAG: {e}")102        return None103 104# Streamlit application105def main():106    st.title("Chat with GitHub Repository")107    st.caption("This app allows you to interact with a GitHub repository using OpenAI and ChromaDB.")108 109    # Get user inputs110    github_token = st.text_input("Enter your GitHub Token", type="password")111    git_repo = st.text_input("Enter the GitHub Repo (owner/repo)")112 113    if github_token and git_repo:114        repo_data = fetch_github_repo_data(git_repo, github_token)115 116        if repo_data:117            st.success(f"Successfully added {git_repo} to the knowledge base!")118 119            question = st.text_input("Ask any question about the repository")120 121            if question:122                answer = perform_rag(repo_data, question)123 124                if answer:125                    st.subheader("Generated Answer:")126                    st.write(answer)127        else:128            st.error("Failed to fetch repository data. Ensure the repository name and token are correct.")129 130if __name__ == "__main__":131    main()132