app.py

import streamlit as st
from PIL import Image
import torch
import easyocr
import numpy as np
import io
from transformers import CLIPModel, CLIPImageProcessor

# ✅ Fix: set_page_config() must be the first Streamlit command
st.set_page_config(page_title="Multimodal AI Assistant", layout="wide")

# ---- Load CLIP Model (Vision Only) ---- #
@st.cache_resource
def load_clip_model():
    model = CLIPModel.from_pretrained(
        "fxmarty/clip-vision-model-tiny", 
        ignore_mismatched_sizes=True  # ✅ Fix size mismatch
    )
    processor = CLIPImageProcessor.from_pretrained("fxmarty/clip-vision-model-tiny")
    return model, processor

model, processor = load_clip_model()

# ---- Load OCR (EasyOCR) ---- #
@st.cache_resource
def load_ocr():
    return easyocr.Reader(['en'])

reader = load_ocr()

# ---- Streamlit UI ---- #
st.title("🖼️ Multimodal AI Assistant")
st.write("Upload an image and ask a question about it!")

# ---- Upload Image ---- #
uploaded_file = st.file_uploader("📤 Upload an image", type=["jpg", "png", "jpeg"])

if uploaded_file is not None:
    # Convert file to image format
    image = Image.open(uploaded_file).convert("RGB")

    # ✅ Fix: use `use_container_width` instead of `use_column_width`
    st.image(image, caption="Uploaded Image", use_container_width=True)

    # ✅ Convert PIL image to NumPy array for EasyOCR
    image_np = np.array(image)

    # ✅ Fix: Pass the correct format to EasyOCR
    with st.spinner("🔍 Extracting text from image..."):
        extracted_text = reader.readtext(image_np, detail=0)

    st.write("### 📝 Extracted Text:")
    if extracted_text:
        st.success(extracted_text)
    else:
        st.warning("No readable text found in the image.")

    # ---- Process Image with CLIP Vision Model ---- #
    with st.spinner("🔍 Analyzing image with CLIP Model..."):
        inputs = processor(images=image, return_tensors="pt")
        outputs = model.get_image_features(**inputs)

    st.write("### 🏆 AI Response:")
    st.write("CLIP Model has processed the image! (More features coming soon)")