Merge branch 'main' of https://github.com/jasonmatthewsuhari/cs3244

Author: jasonmatthewsuhari

README.md

@@ -93,12 +93,12 @@ pip install -r requirements.txt
 ### 5. (Optional) Adding a new model
 If you're adding a new model to the project, please request write access for caching of the model. If not, you can also just add the train code locally
 
-### 6. (Optional) Updatign the report PDF
+### 6. (Optional) Updating the report PDF
 To update the report PDF, you may look into `report/main.tex` and make your updates there, and if you have the `make` tool as well as `pdflatex` installed, you may do the following commands to re-generate the PDF. 
 
 ```bash
 cd reports
 make pdf
 ```
 
-As of right now, the report PDF is in the `.gitignore` to prevent commit clogging, but will be in the final commit for submission. :)
+As of right now, the report PDF is in the `.gitignore` to prevent commit clogging, but will be in the final commit for submission. :)

deployment/chatbot.py

@@ -4,24 +4,41 @@
 from langchain_openai import ChatOpenAI
 from typing import Tuple
 import os
+import numpy
 
-# Rename .env.example to .env, can use bash command below
-# mv .env.example .env
-load_dotenv()
+
+load_dotenv(os.path.join(os.getcwd(), "chatbot", "pyvenv.cfg"))
+api_key = os.getenv("OPENAI_API_KEY")
 
 model = ChatOpenAI(model = "gpt-4o")
 
 process_prompt = PromptTemplate(
-    input_variables = ["text", "has_text", "emoji_img", "emoji_txt"],
+    input_variables = ["text", "image", "has_text", "emoji_img", "emoji_txt"],
+
     template = """
-You are an emoji-enhanced assistant that processes both images and text. Based on the inputs below, produce exactly one output:
+You are an expressive, emoji-savvy assistant that processes both images and text. Based on the inputs below, generate one clear, engaging response:
+
+Instructions:
+- If **has_text** is false:
+  → Generate a fun and creative caption for the image ({image}).
+  → Use the provided emoji ({emoji_img}) as part of the caption.
 
-- If has_text is false: generate an engaging image caption using the emoji from the image ({emoji_img}).
-- If has_text is true and emoji_img is empty: amplify the emotional tone of the text "{text}" using the emoji from text ({emoji_txt}).
-- If both emoji_img and emoji_txt are present: choose the best emoji between {emoji_img} and {emoji_txt}, then amplify "{text}" with that emoji.
+- If **has_text** is true and **emoji_img** is empty:
+  → Enhance the emotional tone of the text: "{text}".
+  → Use the emoji extracted from the text ({emoji_txt}) to enrich it.
 
-Return only the final result and make sure to incorporate the emoji in the text. The emoji doesn't need to be at the end of the text.
+- If both **emoji_img** and **emoji_txt** are present:
+  → Try to guess the caption's emotion and amplify it: "{text}".
+  → Choose the more fitting emoji between {emoji_img} and {emoji_txt}.
+  → Incorporate it to the text that you have amplified!
+
+Final Output:
+Return only the modified caption or text — naturally enhanced with the chosen emoji.
+You can use the emoji more than one time and put it in the middle of the sentences.
+Avoid being excessive on using the emojis as well though.
+Avoid explanations. The output should be concise, expressive, and emoji-infused.
 """
+
 )
 process_chain = LLMChain(llm = model, prompt = process_prompt)
 
@@ -40,6 +57,7 @@ def process(text: str = "", image: Image.Image = None) -> Tuple[str, str]:
 
     response = process_chain.invoke({
         "text": text,
+        "image" : image,
         "has_text": str(has_text).lower(),
         "emoji_img": emoji_img,
         "emoji_txt": emoji_txt,

deployment/streamlit_app.py

@@ -2,9 +2,12 @@
 import streamlit as st
 from PIL import Image
 from chatbot import process
-
+import warnings
 load_dotenv()
 
+warnings.filterwarnings("ignore", category=DeprecationWarning)
+
+
 st.set_page_config(page_title = "Emoji-Enhanced Chatbot", layout = "centered")
 st.title("Emoji-Enhanced Caption & Text Chatbot")
 
@@ -16,13 +19,13 @@
     output, _ = process(user_text, image)
 
     if image and not user_text:
-        st.image(image, caption="Uploaded image", use_column_width=True)
+        st.image(image, caption="Uploaded image", use_container_width=True)
         st.markdown(f"**Caption:** {output}")
     elif user_text and not image:
         st.markdown(f"**Amplified Text:** {output}")
     else:
         if image:
-            st.image(image, caption="Uploaded image", use_column_width=True)
+            st.image(image, caption="Uploaded image", use_container_width=True)
         st.markdown(f"**Output:** {output}")
 else:
     st.info("Please upload an image or enter text, then click Generate.")