added main

app.py

@@ -5,49 +5,62 @@ from PIL import Image
 import os
 
 
-api_key = os.getenv("HUGGINGFACE_TOKEN")
-client = InferenceClient(api_key=api_key)
-
-st.header("Character Captions (IN PROGRESS!)")
-st.write("Have a character caption any image you upload!")
-character = st.selectbox("Choose a character", ["rapper", "shrek", "unintelligible"])
-
-uploaded_img = st.file_uploader("Upload an image")
-
-if uploaded_img is not None:
-
-    image = Image.open(uploaded_img)
-    st.image(image)
-
-    image_captioner = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
-
-    response = image_captioner(image)
-    caption = response[0]['generated_text']
-
-    character_prompts = {
-        "rapper": f"Describe this scene like you're a rapper: {caption}.",
-        "shrek": f"Describe this scene like you're Shrek: {caption}.",
-        "unintelligible": f"Describe this scene in a way that makes no sense: {caption}."
-    }
-
-    prompt = character_prompts[character]
-
-    messages = [
-        { "role": "user", "content": prompt }
-    ]
-
-    stream = client.chat.completions.create(
-        model="meta-llama/Llama-3.2-3B-Instruct", 
-        messages=messages, 
-        max_tokens=500,
-        stream=True
-    )
-
-    response = ''
-    for chunk in stream:
-        response += chunk.choices[0].delta.content
-    
-    st.write(response)
-
-
+def initialize():
+    if 'initialized' not in st.session_state: # Initialize only once
+        print("Initializing...")
+        st.session_state['initialized'] = True
+        st.session_state['api_key'] = os.getenv("HUGGINGFACE_TOKEN")
+        st.session_state['client'] = InferenceClient(api_key=st.session_state['api_key'])
+
+
+def main():
+    initialize()
+    st.header("Character Captions")
+    st.write("Have a character caption any image you upload!")
+    character = st.selectbox("Choose a character", ["rapper", "shrek", "unintelligible", "cookie monster"])
+
+    uploaded_img = st.file_uploader("Upload an image")
+
+    if uploaded_img is not None:
+        # Open Image
+        image = Image.open(uploaded_img)
+        st.image(image)
+
+        # Get caption from image
+        image_captioner = pipeline("image-to-text", model="Salesforce/blip-image-captioning-large")
+        response = image_captioner(image)
+        caption = response[0]['generated_text']
+
+
+        # Pass the caption to a character prompt 
+        character_prompts = {
+            "rapper": f"Describe this caption like you're a rapper: {caption}.",
+            "shrek": f"Describe this caption like you're Shrek: {caption}.",
+            "unintelligible": f"Describe this caption in a way that makes no sense: {caption}.",
+            "cookie monster": f"Describe this caption like you're cookie monster: {caption}."
+        }
+
+        prompt = character_prompts[character]
+        messages = [
+            { "role": "user", "content": prompt }
+        ]
+
+        # Pass to Llama for character output regarding image caption
+        stream = st.session_state['client'].chat.completions.create(
+            model="meta-llama/Llama-3.2-3B-Instruct", 
+            messages=messages, 
+            max_tokens=500,
+            stream=True
+        )
+
+        response = ''
+        for chunk in stream:
+            response += chunk.choices[0].delta.content
+        
+        st.write(response)
+
+
+
+if __name__ == '__main__':
+    main()