import streamlit as st
from huggingface_hub import InferenceClient
st.set_page_config(page_title="JEE AI Mentor", page_icon="🎓", layout="centered")
# --- CUSTOM CSS: GEMINI-INSPIRED MINIMALIST UI ---
st.markdown("""
""", unsafe_allow_html=True)
# --- HEADER ---
st.markdown('JEE Focus AI
', unsafe_allow_html=True)
st.markdown('Your personalized Advanced preparation partner
', unsafe_allow_html=True)
# --- CONTROLS ---
col1, col2 = st.columns(2)
with col1:
target_class = st.selectbox("Class context", ["Class 12 Aspirant", "Dropper Batch", "Class 11"], label_visibility="collapsed")
with col2:
subject = st.selectbox("Subject focus", ["Physics", "Chemistry", "Mathematics"], label_visibility="collapsed")
query_type = st.radio(
"Select operational lens:",
["📚 Doubt", "⏳ Backlog", "🔥 Motivation", "🧠 Revision"],
horizontal=True,
label_visibility="collapsed"
)
# Initialize Chat History
if "messages" not in st.session_state:
st.session_state.messages = [
{"role": "assistant", "content": "Context established. Adjust constraints above whenever you switch tasks. What problem framework or chapter strategy are we evaluating?"}
]
# Display Chat History
for message in st.session_state.messages:
display_content = message["content"]
if "[CONTEXT:" in display_content and "]" in display_content:
display_content = display_content.split("]")[-1].strip()
with st.chat_message(message["role"]):
st.markdown(display_content)
# Input Execution Bar
if user_prompt := st.chat_input("Ask me anything..."):
context_injector = (
f"[CONTEXT: The student is dealing with a {query_type} query for {subject} relevant to {target_class} status. "
"Structure your response specifically through this operational lens.] "
)
with st.chat_message("user"):
st.markdown(user_prompt)
st.session_state.messages.append({"role": "user", "content": context_injector + user_prompt})
with st.chat_message("assistant"):
response_placeholder = st.empty()
full_response = ""
try:
client = InferenceClient()
system_instruction = (
"You are an elite, veteran JEE Advanced mentor who coached top 100 rankers. "
"You understand the exact pressures of Class 11/12/Dropper PCM, standard coaching modules, and PYQ weights.\n"
"CRITICAL: Tailor your response perfectly based on the hidden [CONTEXT] config data:\n"
"- If target is 'Class 12 Aspirant': Balance your tips around board management, school practicals, and running syllabus timelines.\n"
"- If target is 'Dropper Batch': Assume zero school restrictions, focus heavily on maximizing high-yield execution, rigorous test analysis, and handling drop-year mental pressure.\n"
"- If target is 'Class 11': Prioritize fundamental building blocks, tackling initial mechanics/organic shock, and long-term consistency.\n\n"
"Further segment by action lens:\n"
"- If 'Doubt': Break down the core mathematical or conceptual physics/chem mechanisms step-by-step.\n"
"- If 'Backlog': Give a high-yield micro-schedule prioritizing mandatory chapters before deep diving into sub-topics.\n"
"- If 'Motivation': Be blunt, realistic, highly encouraging, and cut through decision paralysis.\n"
"- If 'Revision': Detail how to make high-yield formula sheets and short notes trackers.\n"
"Keep responses concise, bolding critical terms, completely optimized for clean conversational mobile reading. Avoid generic greetings."
)
formatted_messages = [{"role": "system", "content": system_instruction}]
for msg in st.session_state.messages:
formatted_messages.append({"role": msg["role"], "content": msg["content"]})
# Using the ultra-stable, permanently supported serverless model text link
stream = client.chat.completions.create(
model="meta-llama/Meta-Llama-3-8B-Instruct",
messages=formatted_messages,
max_tokens=800,
stream=True
)
for chunk in stream:
if chunk.choices[0].delta.content:
full_response += chunk.choices[0].delta.content
response_placeholder.markdown(full_response + "▌")
response_placeholder.markdown(full_response)
except Exception as e:
st.error(f"Transmission bottleneck: {e}")
full_response = "Hit a brief connection bump. Mind dropping that question in one more time?"
response_placeholder.markdown(full_response)
st.session_state.messages.append({"role": "assistant", "content": full_response})
# --- MINIMALIST FOOTER CREDIT ---
st.markdown("""
""", unsafe_allow_html=True)