AI Debate Partner: Structured Debates With Scoring and Rebuttals

# AI Debate Partner — Streamlit + Groq
# Save as app.py, add GROQ_API_KEY to .streamlit/secrets.toml, then: streamlit run app.py

import ssl

import httpx
import streamlit as st
import truststore

st.set_page_config(page_title="AI Debate Partner", layout="centered")
st.title("AI Debate Partner")
st.caption(
    "Pick a topic, choose a side, and debate an AI that argues its position rigorously. "
    "Challenge its claims and sharpen your critical thinking."
)

_ssl_ctx = truststore.SSLContext(ssl.PROTOCOL_TLS_CLIENT)


@st.cache_resource
def _groq_client():
    import groq

    return groq.Groq(
        api_key=st.secrets["GROQ_API_KEY"],
        http_client=httpx.Client(verify=_ssl_ctx),
    )


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

EXAMPLE_TOPICS = [
    "Remote work is better than office work",
    "AI will eliminate more jobs than it creates",
    "Social media does more harm than good",
    "Universal Basic Income should be implemented",
    "Nuclear energy is the best path to clean energy",
    "Standardized testing should be abolished",
    "Space exploration funding should be prioritized over solving Earth's problems",
]

# ── Setup ──────────────────────────────────────────────────────────────────────
if "debate_active" not in st.session_state:
    st.session_state.debate_active = False
    st.session_state.debate_topic = ""
    st.session_state.ai_side = ""
    st.session_state.user_side = ""
    st.session_state.debate_history = []

if not st.session_state.debate_active:
    st.subheader("Set Up Your Debate")

    topic_choice = st.selectbox("Choose a topic or enter your own", ["Custom…"] + EXAMPLE_TOPICS)
    if topic_choice == "Custom…":
        topic = st.text_input("Your topic", placeholder="e.g. Cats are better pets than dogs")
    else:
        topic = topic_choice

    ai_position = st.radio(
        "The AI will argue:",
        ["FOR (supporting the topic)", "AGAINST (opposing the topic)"],
        horizontal=True,
    )
    ai_side = "FOR" if "FOR" in ai_position else "AGAINST"
    user_side = "AGAINST" if ai_side == "FOR" else "FOR"

    show_both_sides = st.checkbox("Show me both sides first (briefing mode)", value=True)

    start_btn = st.button("Start Debate", type="primary")

    if start_btn:
        if not topic.strip():
            st.warning("Please enter a debate topic.")
        else:
            st.session_state.debate_active = True
            st.session_state.debate_topic = topic
            st.session_state.ai_side = ai_side
            st.session_state.user_side = user_side
            st.session_state.debate_history = []

            if show_both_sides:
                # Generate a briefing on both sides
                system_brief = (
                    "You are a debate coach. Provide a balanced overview of both sides of the topic. "
                    "Format:\n\n"
                    "## FOR (Supporting)\n[3-4 strongest arguments]\n\n"
                    "## AGAINST (Opposing)\n[3-4 strongest arguments]\n\n"
                    "## Key Facts & Statistics\n[relevant data points]\n\n"
                    "Be neutral and academic."
                )
                with st.spinner("Preparing briefing…"):
                    try:
                        resp = _groq_client().chat.completions.create(
                            model=_MODEL,
                            messages=[
                                {"role": "system", "content": system_brief},
                                {"role": "user", "content": f"Topic: {topic}"},
                            ],
                        )
                        briefing = resp.choices[0].message.content
                        st.session_state.debate_history.append(
                            {"role": "assistant", "content": f"**Debate Briefing**\n\n{briefing}"}
                        )
                    except Exception as exc:
                        st.error(f"Error: {exc}")

            # AI opens the debate
            open_system = (
                f"You are a skilled debater arguing {ai_side} the following topic: '{topic}'. "
                "Open the debate with a strong, compelling opening statement (3-4 sentences). "
                "State your position clearly and present your first key argument. "
                "Be persuasive but factual."
            )
            with st.spinner("AI preparing opening statement…"):
                try:
                    resp = _groq_client().chat.completions.create(
                        model=_MODEL,
                        messages=[
                            {"role": "system", "content": open_system},
                            {"role": "user", "content": "Please give your opening statement."},
                        ],
                    )
                    opening = resp.choices[0].message.content
                    st.session_state.debate_history.append(
                        {"role": "assistant", "content": opening}
                    )
                except Exception as exc:
                    st.error(f"Error: {exc}")

            st.rerun()

# ── Active debate ──────────────────────────────────────────────────────────────
if st.session_state.debate_active:
    topic = st.session_state.debate_topic
    ai_side = st.session_state.ai_side
    user_side = st.session_state.user_side

    st.info(
        f"**Topic:** {topic}  \n"
        f"**AI argues:** {ai_side}  |  **You argue:** {user_side}"
    )

    for msg in st.session_state.debate_history:
        role = msg["role"]
        with st.chat_message(role):
            st.markdown(msg["content"])

    user_arg = st.chat_input(f"Make your argument ({user_side})…")

    if user_arg:
        st.session_state.debate_history.append({"role": "user", "content": user_arg})
        with st.chat_message("user"):
            st.markdown(user_arg)

        system_msg = (
            f"You are a skilled debater arguing {ai_side} the topic: '{topic}'. "
            f"The human is arguing {user_side}. "
            "Respond to their argument by:\n"
            "1. Acknowledging what they said (briefly)\n"
            "2. Pointing out a weakness or flaw in their argument\n"
            "3. Reinforcing your own position with a new supporting point or evidence\n\n"
            "Be intellectually rigorous, respectful, and persuasive. "
            "Keep your response to 3-5 sentences. Stay in character as the debater."
        )

        messages = [{"role": "system", "content": system_msg}]
        for h in st.session_state.debate_history[-8:]:
            messages.append(h)

        with st.chat_message("assistant"):
            with st.spinner("AI responding…"):
                try:
                    resp = _groq_client().chat.completions.create(
                        model=_MODEL,
                        messages=messages,
                    )
                    ai_response = resp.choices[0].message.content
                    st.markdown(ai_response)
                    st.session_state.debate_history.append(
                        {"role": "assistant", "content": ai_response}
                    )
                except Exception as exc:
                    st.error(f"Error: {exc}")

    st.divider()
    col1, col2 = st.columns(2)

    with col1:
        if st.button("Request Verdict", use_container_width=True):
            verdict_system = (
                "You are an impartial debate judge. Review the debate and provide:\n\n"
                "## Verdict\n[Who made stronger arguments overall]\n\n"
                "## Scoring\n- Argument strength: [X/10 for each side]\n"
                "- Use of evidence: [X/10 for each side]\n"
                "- Rebuttal quality: [X/10 for each side]\n\n"
                "## Key Moments\n[Most compelling points from each side]\n\n"
                "## What Could Improve\n[One tip for each side]\n\n"
                "Be fair and analytical."
            )
            debate_text = "\n\n".join(
                f"{'AI' if m['role'] == 'assistant' else 'Human'}: {m['content']}"
                for m in st.session_state.debate_history
            )
            with st.spinner("Judging…"):
                try:
                    resp = _groq_client().chat.completions.create(
                        model=_MODEL,
                        messages=[
                            {"role": "system", "content": verdict_system},
                            {"role": "user", "content": f"Debate transcript:\n{debate_text[:8000]}"},
                        ],
                    )
                    st.markdown(resp.choices[0].message.content)
                except Exception as exc:
                    st.error(f"Error: {exc}")

    with col2:
        if st.button("Start New Debate", use_container_width=True):
            st.session_state.debate_active = False
            st.session_state.debate_history = []
            st.rerun()