Update app.py

app.py

@@ -1,277 +1,41 @@
 import streamlit as st
-import pandas as pd
-import google.generativeai as genai
-import re
-from PIL import Image
+from transformers import AutoTokenizer, AutoModelForImageCaptioning
 import requests
-
-#Je t'aime plus que les mots,
-#Plus que les sentiments,
-#Plus que la vie elle-même
-
-st.set_page_config(
-    page_title="Google AI Chat",
-    page_icon="https://seeklogo.com/images/G/google-ai-logo-996E85F6FD-seeklogo.com.png",
-    layout="wide",
-)
-# Path: Main.py
-#Author: Sergio Demis Lopez Martinez
-#------------------------------------------------------------
-#HEADER
-st.markdown('''
-Powered by Google AI <img src="https://seeklogo.com/images/G/google-ai-logo-996E85F6FD-seeklogo.com.png" width="20" height="20">
-, Streamlit and Python''', unsafe_allow_html=True)
-st.caption("By Sergio Demis Lopez Martinez")
-
-#------------------------------------------------------------
-#LANGUAGE
-langcols = st.columns([0.2,0.8])
-with langcols[0]:
-  lang = st.selectbox('Select your language',
-  ('English', 'Español', 'Français', 'Deutsch',
-  'Italiano', 'Português', 'Polski', 'Nederlands',
-  'Русский', '日本語', '한국어', '中文', 'العربية',
-  'हिन्दी', 'Türkçe', 'Tiếng Việt', 'Bahasa Indonesia',
-  'ภาษาไทย', 'Română', 'Ελληνικά', 'Magyar', 'Čeština',
-  'Svenska', 'Norsk', 'Suomi', 'Dansk', 'हिन्दी', 'हिन्�'),index=1)
-
-if 'lang' not in st.session_state:
-    st.session_state.lang = lang
-st.divider()
-
-#------------------------------------------------------------
-#FUNCTIONS
-def extract_graphviz_info(text: str) -> list[str]:
-  """
-  The function `extract_graphviz_info` takes in a text and returns a list of graphviz code blocks found in the text.
-
-  :param text: The `text` parameter is a string that contains the text from which you want to extract Graphviz information
-  :return: a list of strings that contain either the word "graph" or "digraph". These strings are extracted from the input
-  text.
-  """
-
-  graphviz_info  = text.split('```')
-
-  return [graph for graph in graphviz_info if ('graph' in graph or 'digraph' in graph) and ('{' in graph and '}' in graph)]
-
-def append_message(message: dict) -> None:
-    """
-    The function appends a message to a chat session.
-
-    :param message: The `message` parameter is a dictionary that represents a chat message. It typically contains
-    information such as the user who sent the message and the content of the message
-    :type message: dict
-    :return: The function is not returning anything.
-    """
-    st.session_state.chat_session.append({'user': message})
-    return
-
-@st.cache_resource
-def load_model() -> genai.GenerativeModel:
-    """
-    The function `load_model()` returns an instance of the `genai.GenerativeModel` class initialized with the model name
-    'gemini-pro'.
-    :return: an instance of the `genai.GenerativeModel` class.
-    """
-    model = genai.GenerativeModel('gemini-pro')
-    return model
-
-@st.cache_resource
-def load_modelvision() -> genai.GenerativeModel:
-    """
-    The function `load_modelvision` loads a generative model for vision tasks using the `gemini-pro-vision` model.
-    :return: an instance of the `genai.GenerativeModel` class.
-    """
-    model = genai.GenerativeModel('gemini-pro-vision')
-    return model
-
-
-
-#------------------------------------------------------------
-#CONFIGURATION
-genai.configure(api_key=st.secrets["GOOGLE_API_KEY"])
-
-model = load_model()
-
-vision = load_modelvision()
-
-if 'chat' not in st.session_state:
-    st.session_state.chat = model.start_chat(history=[])
-
-if 'chat_session' not in st.session_state:
-    st.session_state.chat_session = []
-
-#st.session_state.chat_session
-
-#------------------------------------------------------------
-#CHAT
-
-if 'messages' not in st.session_state:
-    st.session_state.messages = []
-
-if 'welcome' not in st.session_state or lang != st.session_state.lang:
-    st.session_state.lang = lang
-    welcome  = model.generate_content(f'''
-    Da un saludo de bienvenida al usuario y sugiere que puede hacer
-    (Puedes describir imágenes, responder preguntas, leer archivos texto, leer tablas,generar gráficos con graphviz, etc)
-    eres un chatbot en una aplicación de chat creada en streamlit y python. generate the answer in {lang}''')
-    welcome.resolve()
-    st.session_state.welcome = welcome
-
-    with st.chat_message('ai'):
-        st.write(st.session_state.welcome.text)
-else:
-    with st.chat_message('ai'):
-        st.write(st.session_state.welcome.text)
-
-if len(st.session_state.chat_session) > 0:
-    count = 0
-    for message in st.session_state.chat_session:
-
-        if message['user']['role'] == 'model':
-            with st.chat_message('ai'):
-                st.write(message['user']['parts'])
-                graphs = extract_graphviz_info(message['user']['parts'])
-                if len(graphs) > 0:
-                    for graph in graphs:
-                        st.graphviz_chart(graph,use_container_width=False)
-                        if lang == 'Español':
-                          view = "Ver texto"
-                        else:
-                          view = "View text"
-                        with st.expander(view):
-                          st.code(graph, language='dot')
-        else:
-            with st.chat_message('user'):
-                st.write(message['user']['parts'][0])
-                if len(message['user']['parts']) > 1:
-                    st.image(message['user']['parts'][1], width=200)
-        count += 1
-
-
-
-#st.session_state.chat.history
-
-cols=st.columns(4)
-
-with cols[0]:
-    if lang == 'Español':
-      image_atachment = st.toggle("Adjuntar imagen", value=False, help="Activa este modo para adjuntar una imagen y que el chatbot pueda leerla")
-    else:
-      image_atachment = st.toggle("Attach image", value=False, help="Activate this mode to attach an image and let the chatbot read it")
-
-with cols[1]:
-    if lang == 'Español':
-      txt_atachment = st.toggle("Adjuntar archivo de texto", value=False, help="Activa este modo para adjuntar un archivo de texto y que el chatbot pueda leerlo")
-    else:
-      txt_atachment = st.toggle("Attach text file", value=False, help="Activate this mode to attach a text file and let the chatbot read it")
-with cols[2]:
-    if lang == 'Español':
-      csv_excel_atachment = st.toggle("Adjuntar CSV o Excel", value=False, help="Activa este modo para adjuntar un archivo CSV o Excel y que el chatbot pueda leerlo")
-    else:
-      csv_excel_atachment = st.toggle("Attach CSV or Excel", value=False, help="Activate this mode to attach a CSV or Excel file and let the chatbot read it")
-with cols[3]:
-    if lang == 'Español':
-      graphviz_mode = st.toggle("Modo graphviz", value=False, help="Activa este modo para generar un grafo con graphviz en .dot a partir de tu mensaje")
-    else:
-      graphviz_mode = st.toggle("Graphviz mode", value=False, help="Activate this mode to generate a graph with graphviz in .dot from your message")
-if image_atachment:
-    if lang == 'Español':
-      image = st.file_uploader("Sube tu imagen", type=['png', 'jpg', 'jpeg'])
-      url = st.text_input("O pega la url de tu imagen")
-    else:
-      image = st.file_uploader("Upload your image", type=['png', 'jpg', 'jpeg'])
-      url = st.text_input("Or paste your image url")
-else:
-    image = None
-    url = ''
-
-
-
-if txt_atachment:
-    if lang == 'Español':
-      txtattachment = st.file_uploader("Sube tu archivo de texto", type=['txt'])
-    else:
-      txtattachment = st.file_uploader("Upload your text file", type=['txt'])
-else:
-    txtattachment = None
-
-if csv_excel_atachment:
-    if lang == 'Español':
-      csvexcelattachment = st.file_uploader("Sube tu archivo CSV o Excel", type=['csv', 'xlsx'])
-    else:
-      csvexcelattachment = st.file_uploader("Upload your CSV or Excel file", type=['csv', 'xlsx'])
-else:
-    csvexcelattachment = None
-if lang == 'Español':
-  prompt = st.chat_input("Escribe tu mensaje")
-else:
-  prompt = st.chat_input("Write your message")
-
-if prompt:
-    txt = ''
-    if txtattachment:
-        txt = txtattachment.getvalue().decode("utf-8")
-        if lang == 'Español':
-          txt = '   Archivo de texto: \n' + txt
-        else:
-          txt = '   Text file: \n' + txt
-
-    if csvexcelattachment:
-        try:
-            df = pd.read_csv(csvexcelattachment)
-        except:
-            df = pd.read_excel(csvexcelattachment)
-        txt += '   Dataframe: \n' + str(df)
-
-    if graphviz_mode:
-        if lang == 'Español':
-          txt += '   Genera un grafo con graphviz en .dot \n'
-        else:
-          txt += '   Generate a graph with graphviz in .dot \n'
-
-    if len(txt) > 5000:
-        txt = txt[:5000] + '...'
-    if image or url != '':
-        if url != '':
-            img = Image.open(requests.get(url, stream=True).raw)
-        else:
-            img = Image.open(image)
-        prmt  = {'role': 'user', 'parts':[prompt+txt, img]}
-    else:
-        prmt  = {'role': 'user', 'parts':[prompt+txt]}
-
-    append_message(prmt)
-
-    if lang == 'Español':
-      spinertxt = 'Espera un momento, estoy pensando...'
-    else:
-      spinertxt = 'Wait a moment, I am thinking...'
-    with st.spinner(spinertxt):
-        if len(prmt['parts']) > 1:
-            response = vision.generate_content(prmt['parts'],stream=True,safety_settings=[
-        {
-            "category": "HARM_CATEGORY_HARASSMENT",
-            "threshold": "BLOCK_LOW_AND_ABOVE",
-        },
-        {
-            "category": "HARM_CATEGORY_HATE_SPEECH",
-            "threshold": "BLOCK_LOW_AND_ABOVE",
-        },
-    ]
-)
-            response.resolve()
-        else:
-            response = st.session_state.chat.send_message(prmt['parts'][0])
-
-        try:
-          append_message({'role': 'model', 'parts':response.text})
-        except Exception as e:
-          append_message({'role': 'model', 'parts':f'{type(e).__name__}: {e}'})
-
-
-        st.rerun()
-
-
-
-#st.session_state.chat_session
+from PIL import Image
+import numpy as np
+
+# Initialize the tokenizer and model
+tokenizer = AutoTokenizer.from_pretrained("microsoft/beit-base-patch16-224-in21k")
+model = AutoModelForImageCaptioning.from_pretrained("microsoft/beit-base-patch16-224-in21k")
+
+def generate_caption(image_url):
+  # Get the image from the URL
+  image = Image.open(requests.get(image_url, stream=True).raw)
+  
+  # Preprocess the image
+  input_array = np.array(image) / 255.0
+  input_array = np.transpose(input_array, (2, 0, 1))
+  input_ids = tokenizer(image_url, return_tensors="pt").input_ids
+
+  # Generate the caption
+  output = model.generate(input_ids, max_length=20)
+  caption = tokenizer.batch_decode(output, skip_special_tokens=True)
+
+  return caption[0]
+
+def main():
+  # Create a sidebar for the user to input the image URL
+  st.sidebar.header("Image Caption Generator")
+  image_url = st.sidebar.text_input("Enter the URL of an image:")
+
+  # Generate the caption if the user clicks the button
+  if st.sidebar.button("Generate Caption"):
+    if image_url != "":
+      caption = generate_caption(image_url)
+      st.success(f"Caption: {caption}")
+    else:
+      st.error("Please enter a valid image URL.")
+
+# Run the main function
+if __name__ == "__main__":
+    main()