import streamlit as st
from dotenv import load_dotenv
import os
import tempfile
import base64
from gtts import gTTS # ✅ Use gTTS instead of pyttsx3
from langchain_groq import ChatGroq
from langchain.chains import LLMChain
from langchain.prompts import PromptTemplate
# Load environment variables
load_dotenv()
GROQ_API_KEY = os.getenv("GROQ_API_KEY")
# ✅ Exact optimized prompt as provided
optimized_prompt = PromptTemplate(
input_variables=["problem"],
template=(
"You are an advanced AI tutor specializing in solving **math and physics problems** step by step. "
"Your goal is to **guide students logically**, ensuring they **understand every step** and its relevance. "
"Think like a **patient teacher** who explains concepts with clarity.\n\n"
"📚 **Guidelines for solving problems:**\n"
"1️⃣ **Understanding the Problem:**\n"
" - Restate the problem in simple terms.\n"
" - Identify what is given and what needs to be found.\n\n"
"2️⃣ **Relevant Concepts & Formulas:**\n"
" - List the key principles, equations, or theorems needed to solve the problem.\n"
" - Explain why they are relevant.\n\n"
"3️⃣ **Step-by-Step Solution:**\n"
" - Break down the solution into small, logical steps.\n"
" - Show calculations with proper notation.\n"
" - Explain **each transformation, substitution, or simplification** clearly.\n\n"
"4️⃣ **Final Answer:**\n"
" - ✅ Box or highlight the final result.\n"
" - Include units where applicable.\n\n"
"5️⃣ **Verification & Insights:**\n"
" - 🔄 Verify the answer using an alternative method (if possible).\n"
" - 🏗️ Provide a real-world analogy or intuition behind the result.\n\n"
"🎯 **Now, solve the following problem using this structured approach:**\n"
"**Problem:** {problem}\n\n"
"**Solution:**"
)
)
# Initialize Groq API model
llm = ChatGroq(api_key=GROQ_API_KEY, model_name="gemma2-9b-it")
llm_chain = LLMChain(llm=llm, prompt=optimized_prompt)
# Function to generate speech using gTTS and return Base64-encoded audio
def text_to_speech(text):
"""Generate gTTS audio and return Base64 encoded string."""
tts = gTTS(text, lang="en")
with tempfile.NamedTemporaryFile(delete=False, suffix=".mp3") as temp_audio:
tts.save(temp_audio.name)
with open(temp_audio.name, "rb") as audio_file:
audio_bytes = audio_file.read()
encoded_audio = base64.b64encode(audio_bytes).decode()
os.remove(temp_audio.name)
return encoded_audio
# Streamlit UI Design
st.set_page_config(page_title="STEM Solver 🤖", layout="centered", page_icon="🧠")
st.markdown("
📚 STEM Problem Solver 🤖
", unsafe_allow_html=True)
st.markdown("
Enter a math or physics problem, and I'll solve it step by step! 🚀
", unsafe_allow_html=True)
# User Input
problem = st.text_area("📝 Enter your problem:", placeholder="e.g., What is the integral of x?", height=100)
# Solve (Text) Button
if st.button("🔍 Solve (Text)"):
if problem.strip():
with st.spinner("Thinking... 🤔"):
response = llm_chain.invoke({"problem": problem})
solution_text = response['text']
st.success("✅ Solution Found!")
st.markdown("### ✨ Solution:")
st.markdown(f"
"
f"
{solution_text}
", unsafe_allow_html=True)
else:
st.warning("⚠️ Please enter a valid problem.")
# Solve (Speech) Button
if st.button("🔊 Solve (Speech)"):
if problem.strip():
with st.spinner("Speaking... 🎤"):
response = llm_chain.invoke({"problem": problem})
solution_text = response['text']
audio_base64 = text_to_speech(solution_text) # Get Base64 audio
audio_html = f"""
"""
st.success("✅ Solve (Speech) Started!")
st.markdown(audio_html, unsafe_allow_html=True)
# 📥 Add a download button for mobile users
audio_file_name = "solution.mp3"
with open(audio_file_name, "wb") as file:
file.write(base64.b64decode(audio_base64))
with open(audio_file_name, "rb") as file:
st.download_button(
label="📥 Download Audio",
data=file,
file_name="solution.mp3",
mime="audio/mp3"
)
else:
st.warning("⚠️ Please enter a valid problem.")
# Footer
st.markdown("