src/streamlit_app.py

import streamlit as st
import torch
import random
from transformers import AutoTokenizer, AutoModelForSeq2SeqLM

# ---------------- PAGE CONFIG ----------------
st.set_page_config(page_title="FitPlan AI", page_icon="💪", layout="centered")

# ---------------- SESSION STATE ----------------
if "page" not in st.session_state:
    st.session_state.page = "landing"

# ---------------- LANDING PAGE ----------------
if st.session_state.page == "landing":

    st.markdown("""
        <style>
        [data-testid="stAppViewContainer"] {
            background-image: url("https://images.unsplash.com/photo-1483721310020-03333e577078");
            background-size: cover;
            background-position: center;
        }
        .center-box {
            text-align:center;
            margin-top:200px;
            font-size:48px;
            font-weight:bold;
            color:black;
        }
        </style>
    """, unsafe_allow_html=True)

    st.markdown('<div class="center-box">💪 Welcome to FitPlan AI</div>', unsafe_allow_html=True)

    if st.button("🚀 Get Started"):
        st.session_state.page = "main"
        st.rerun()

# ---------------- MAIN PAGE ----------------
elif st.session_state.page == "main":

    st.title("💪 FitPlan AI")

    # ---------------- INPUTS ----------------
    name = st.text_input("Name *")
    gender = st.selectbox("Gender", ["Male", "Female", "Other"])
    height_cm = st.number_input("Height (cm) *", min_value=0.0)
    weight_kg = st.number_input("Weight (kg) *", min_value=0.0)

    goal = st.selectbox(
        "Fitness Goal",
        ["Build Muscle", "Weight Loss", "Strength Gain", "Abs Building", "Flexible"]
    )

    equipment = st.multiselect(
        "Equipment",
        ["Dumbbells", "Resistance Band", "Yoga Mat", "No Equipment"]
    )

    fitness_level = st.radio(
        "Fitness Level",
        ["Beginner", "Intermediate", "Advanced"]
    )

    # ---------------- BMI FUNCTIONS ----------------
    def calculate_bmi(h, w):
        return round(w / ((h / 100) ** 2), 2)

    def bmi_category(b):
        if b < 18.5:
            return "Underweight"
        elif b < 25:
            return "Normal"
        elif b < 30:
            return "Overweight"
        else:
            return "Obese"

    # ---------------- BMI BUTTON ----------------
    if st.button("Generate BMI"):
        if height_cm <= 0 or weight_kg <= 0:
            st.error("Enter valid height and weight.")
        else:
            bmi_value = calculate_bmi(height_cm, weight_kg)
            st.session_state["bmi"] = bmi_value
            st.success(f"BMI: {bmi_value:.2f} ({bmi_category(bmi_value)})")

    bmi = st.session_state.get("bmi", None)

    # ---------------- MODEL LOADING ----------------
    @st.cache_resource
    def load_model():
        tokenizer = AutoTokenizer.from_pretrained("google/flan-t5-base")
        model = AutoModelForSeq2SeqLM.from_pretrained("google/flan-t5-base")
        return tokenizer, model

    tokenizer, model = load_model()

    # ---------------- SUBMIT PROFILE ----------------
    if st.button("Submit Profile"):

        if not name:
            st.error("Enter your name.")
        elif height_cm <= 0 or weight_kg <= 0:
            st.error("Enter valid height & weight.")
        elif not equipment:
            st.error("Select equipment.")
        elif bmi is None:
            st.error("Generate BMI first.")
        else:
            st.success("Profile Submitted Successfully!")

            bmi_status = bmi_category(bmi)
            equipment_list = ", ".join(equipment)

            # Add randomness token to force variation
            random_token = random.randint(1, 1000000)

            prompt = f"""
You are a certified professional fitness trainer.

Random Seed: {random_token}

Generate a structured 5-day workout plan based on the following user profile.

User Profile:
- Name: {name}
- Gender: {gender}
- BMI: {bmi:.2f} ({bmi_status})
- Goal: {goal}
- Fitness Level: {fitness_level}
- Available Equipment: {equipment_list}

Instructions:
1. Divide the plan clearly into Day 1 to Day 5.
2. Under each day, list 4-6 exercises.
3. For each exercise include:
   - Exercise Name
   - Sets
   - Reps
   - Rest Time
4. Keep exercises appropriate for fitness level.
5. Do NOT include explanations outside workout plan.

Only return the workout plan.
"""

            with st.spinner("Generating Workout Plan..."):

                inputs = tokenizer(prompt, return_tensors="pt", truncation=True)

                outputs = model.generate(
                    **inputs,
                    max_new_tokens=400,
                    do_sample=True,
                    temperature=0.9,        # Increased randomness
                    top_p=0.95,
                    top_k=50,
                    repetition_penalty=1.5,
                    num_beams=1             # Important: disables deterministic beam search
                )

                result = tokenizer.decode(outputs[0], skip_special_tokens=True).strip()

            st.subheader("🏋️ Your Personalized Workout Plan")
            st.write(result)