Personal Study Tutor: A RAG Tutor That Reads Your PDFs and Quizzes You

# Personal Study Tutor (RAG-based) — Streamlit + LangChain + Chroma + Groq
# Save as app.py, add GROQ_API_KEY to .streamlit/secrets.toml, then: streamlit run app.py

import io
import ssl

import httpx
import streamlit as st
import truststore

MAX_NOTES_CHARS = 12000
MAX_PDF_CHARS = 15000

st.set_page_config(page_title="Personal Study Tutor", layout="centered")
st.title("Personal Study Tutor")
st.caption(
    "Upload your notes or PDFs, build a knowledge index, then ask questions "
    "or request a quiz on any topic inside your material."
)

_ssl_ctx = truststore.SSLContext(ssl.PROTOCOL_TLS_CLIENT)


@st.cache_resource
def _groq_http_client():
    return httpx.Client(verify=_ssl_ctx)


def _get_groq_client():
    import groq

    if "groq_client" not in st.session_state:
        st.session_state.groq_client = groq.Groq(
            api_key=st.secrets["GROQ_API_KEY"],
            http_client=_groq_http_client(),
        )
    return st.session_state.groq_client


_CHAT_MODEL = "llama-3.3-70b-versatile"

# ── Sidebar: input source ──────────────────────────────────────────────────────
with st.sidebar:
    st.header("Your Study Material")
    uploaded_files = st.file_uploader(
        "Upload PDFs or TXT files",
        type=["pdf", "txt"],
        accept_multiple_files=True,
    )
    pasted_notes = st.text_area(
        "Or paste notes here",
        height=200,
        placeholder="Paste any text you want to study...",
    )
    build_btn = st.button("Build / Refresh Index", type="primary", use_container_width=True)

# ── Build index ────────────────────────────────────────────────────────────────
if build_btn:
    from langchain.schema import Document
    from langchain_community.vectorstores import Chroma
    from langchain_huggingface import HuggingFaceEmbeddings
    from langchain_text_splitters import RecursiveCharacterTextSplitter

    all_text = ""

    for f in (uploaded_files or []):
        if f.name.endswith(".pdf"):
            import pypdf

            reader = pypdf.PdfReader(io.BytesIO(f.read()))
            for page in reader.pages:
                t = page.extract_text() or ""
                all_text += t + "\n"
        else:
            all_text += f.read().decode("utf-8", errors="replace") + "\n"

    if pasted_notes and pasted_notes.strip():
        all_text += pasted_notes.strip()

    all_text = all_text[:MAX_PDF_CHARS]

    if not all_text.strip():
        st.sidebar.warning("Please upload a file or paste some notes first.")
    else:
        with st.sidebar:
            with st.spinner("Building index…"):
                try:
                    splitter = RecursiveCharacterTextSplitter(chunk_size=600, chunk_overlap=80)
                    docs = splitter.split_documents([Document(page_content=all_text)])
                    embeddings = HuggingFaceEmbeddings(
                        model_name="sentence-transformers/all-MiniLM-L6-v2"
                    )
                    st.session_state.vectorstore = Chroma.from_documents(docs, embeddings)
                    st.session_state.doc_count = len(docs)
                    st.success(f"Index built — {len(docs)} chunks ready.")
                except Exception as exc:
                    st.error(f"Index error: {exc}")

if "vectorstore" in st.session_state:
    st.sidebar.caption(f"Index: {st.session_state.doc_count} chunks loaded.")

# ── Main area ──────────────────────────────────────────────────────────────────
tab_ask, tab_quiz = st.tabs(["Ask a Question", "Generate Quiz"])

# ── Tab 1: Ask ─────────────────────────────────────────────────────────────────
with tab_ask:
    st.subheader("Ask Anything From Your Notes")
    question = st.text_input(
        "Your question",
        placeholder="What is the main argument in Chapter 3?",
    )
    ask_btn = st.button("Get Answer", key="ask_btn")

    if ask_btn:
        if "vectorstore" not in st.session_state:
            st.warning("Build the index first using the sidebar.")
        elif not question.strip():
            st.warning("Enter a question.")
        else:
            from langchain_groq import ChatGroq
            from langchain.chains import RetrievalQA

            with st.spinner("Searching notes and generating answer…"):
                try:
                    retriever = st.session_state.vectorstore.as_retriever(
                        search_kwargs={"k": 4}
                    )
                    llm = ChatGroq(
                        model_name=_CHAT_MODEL,
                        groq_api_key=st.secrets["GROQ_API_KEY"],
                        http_client=_groq_http_client(),
                    )
                    qa = RetrievalQA.from_chain_type(
                        llm=llm,
                        retriever=retriever,
                        return_source_documents=True,
                        chain_type="stuff",
                    )
                    result = qa.invoke({"query": question})
                    st.markdown("### Answer")
                    st.markdown(result["result"])
                    with st.expander("Source chunks used"):
                        for i, doc in enumerate(result["source_documents"], 1):
                            st.markdown(f"**Chunk {i}:** {doc.page_content[:300]}…")
                except Exception as exc:
                    st.error(f"Error: {exc}")

# ── Tab 2: Quiz ────────────────────────────────────────────────────────────────
with tab_quiz:
    st.subheader("Test Yourself")
    col1, col2 = st.columns(2)
    with col1:
        topic_focus = st.text_input(
            "Topic or chapter focus (optional)",
            placeholder="e.g. photosynthesis, Chapter 2",
        )
    with col2:
        num_questions = st.selectbox("Number of questions", [3, 5, 10], index=1)

    quiz_btn = st.button("Generate Quiz", key="quiz_btn")

    if quiz_btn:
        if "vectorstore" not in st.session_state:
            st.warning("Build the index first using the sidebar.")
        else:
            context_text = ""
            if topic_focus.strip():
                try:
                    retriever = st.session_state.vectorstore.as_retriever(
                        search_kwargs={"k": 5}
                    )
                    docs = retriever.invoke(topic_focus)
                    context_text = "\n\n".join(d.page_content for d in docs)
                except Exception:
                    context_text = ""

            if not context_text:
                # Fall back to raw notes if retrieval fails
                context_text = st.session_state.get("raw_text", "")[:3000]

            if not context_text:
                st.warning("Could not retrieve relevant content. Try rebuilding the index.")
            else:
                system_msg = (
                    "You are a study tutor. Create a multiple-choice quiz with "
                    f"{num_questions} questions based on the provided study material. "
                    "Format each question as:\n"
                    "Q1. [Question]\n"
                    "A) ...\nB) ...\nC) ...\nD) ...\n"
                    "Answer: [letter]\nExplanation: [one sentence]\n\n"
                    "Focus only on content that appears in the material."
                )
                user_msg = f"Study material:\n{context_text[:4000]}"
                if topic_focus.strip():
                    user_msg += f"\n\nFocus specifically on: {topic_focus}"

                with st.spinner("Generating quiz…"):
                    try:
                        client = _get_groq_client()
                        resp = client.chat.completions.create(
                            model=_CHAT_MODEL,
                            messages=[
                                {"role": "system", "content": system_msg},
                                {"role": "user", "content": user_msg},
                            ],
                        )
                        st.markdown("### Your Quiz")
                        st.markdown(resp.choices[0].message.content)
                    except Exception as exc:
                        st.error(f"Error: {exc}")