import gradio as gr
from transformers import CLIPProcessor, CLIPModel, CLIPTokenizer
import sentence_transformers
from sentence_transformers import SentenceTransformer, util
import pickle
from PIL import Image
import os


## Define model
model = CLIPModel.from_pretrained("openai/clip-vit-base-patch32")
processor = CLIPProcessor.from_pretrained("openai/clip-vit-base-patch32")
tokenizer = CLIPTokenizer.from_pretrained("openai/clip-vit-base-patch32")

#Open the precomputed embeddings
emb_filename = 'unsplash-25k-photos-embeddings.pkl'
with open(emb_filename, 'rb') as fIn:
        img_names, img_emb = pickle.load(fIn)
        #print(f'img_emb: {print(img_emb)}')
        #print(f'img_names: {print(img_names)}')  


def search_text(query, top_k=1):
    """" Search an image based on the text query.
    
    Args:
        query ([string]): [query you want search for]
        top_k (int, optional): [Amount of images o return]. Defaults to 1.

    Returns:
        [list]: [list of images that are related to the query.]
    """
    # First, we encode the query.
    inputs = tokenizer([query],  padding=True, return_tensors="pt")
    query_emb = model.get_text_features(**inputs)
    
    # Then, we use the util.semantic_search function, which computes the cosine-similarity
    # between the query embedding and all image embeddings.
    # It then returns the top_k highest ranked images, which we output
    hits = util.semantic_search(query_emb, img_emb, top_k=top_k)[0]
    
    image=[]
    for hit in hits:
        #print(img_names[hit['corpus_id']])
        object = Image.open(os.path.join("photos/", img_names[hit['corpus_id']]))
        image.append(object)
        #print(f'array length is: {len(image)}')

    return image
    

iface = gr.Interface(
    title = "Text to Image using CLIP Model 📸",
    description = "Gradio Demo fo CLIP model. \n This demo is based on assessment for the 🤗  Huggingface course 2. \n To use it, simply write which image you are looking for. Read more at the links below.",
    article = "You find more information about this demo on my ✨ github repository [marcelcastrobr](https://github.com/marcelcastrobr/huggingface_course2)",
    fn=search_text, 
    inputs=[gr.Textbox(lines=4,
        label="Write what you are looking for in an image...",
        placeholder="Text Here..."),
        gr.Slider(0, 5, step=1)],
    outputs=[gr.Gallery(
            label="Generated images", show_label=False, elem_id="gallery"
        ).style(grid=[2], height="auto")]
    ,examples=[[("Dog in the beach"), 2], 
        [("Paris during night."), 1], 
        [("A cute kangaroo"), 5],
        [("Dois cachorros"), 2],
        [("un homme marchant sur le parc"), 3],
        [("et høyt fjell"), 2]]
    ).launch(debug=True)