Update app.py

app.py

@@ -1,139 +1,83 @@
-# app.py
 import streamlit as st
-from langchain.embeddings import HuggingFaceEmbeddings
-from langchain.vectorstores import FAISS
-from transformers import AutoTokenizer, AutoModelForCausalLM, pipeline
-import torch
-import random
+from textblob import TextBlob
+# The transformers import is here in case you want to integrate a pre-trained model later.
+# from transformers import pipeline
 
-# --------------------- Configuration ---------------------
-MODEL_NAME = "mistralai/Mistral-7B-Instruct-v0.2"
-EMBEDDING_MODEL = "sentence-transformers/all-mpnet-base-v2"
-CONSPIRACY_DB_PATH = "conspiracy_faiss_index"
-FACT_DB_PATH = "facts_faiss_index"
+def get_roast(user_text):
+    """
+    Generates a roast based on the user's input.
+    (You can replace or extend this logic by integrating a model like Mixtral-8x7B.)
+    """
+    if "procrastinate" in user_text.lower():
+        return "Ah, the art of doing nothing! Do you charge Netflix for your couch imprint? 🛋️"
+    elif "always" in user_text.lower() or "never" in user_text.lower():
+        return "Wow, painting your world in extremes? Maybe it's time to add some shades of gray!"
+    else:
+        return "Is that a problem or a lifestyle choice? Time to get serious... or maybe not."
 
-# --------------------- Helper Functions ---------------------
-def load_vector_db(path):
-    embeddings = HuggingFaceEmbeddings(model_name=EMBEDDING_MODEL)
-    return FAISS.load_local(path, embeddings)
+def analyze_text(user_text):
+    """
+    Analyzes the input text for sentiment and detects basic cognitive distortions.
+    For now, it counts occurrences of words like 'always' or 'never'.
+    """
+    blob = TextBlob(user_text)
+    sentiment = blob.sentiment.polarity  # Ranges from -1 (negative) to 1 (positive)
+    distortions = sum(word in user_text.lower() for word in ["always", "never"])
+    return sentiment, distortions
 
-def load_llm():
-    tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
-    model = AutoModelForCausalLM.from_pretrained(
-        MODEL_NAME,
-        device_map="auto",
-        load_in_4bit=True,
-        torch_dtype=torch.float16
-    )
-    return pipeline(
-        "text-generation",
-        model=model,
-        tokenizer=tokenizer,
-        device_map="auto"
-    )
+def calculate_resilience_score(sentiment, distortions):
+    """
+    Calculates a resilience score based on sentiment and the number of detected distortions.
+    The score is capped between 0 and 100.
+    """
+    score = 100
+    # Adjust score by sentiment (scaled)
+    score += int(sentiment * 20)
+    # Penalize for cognitive distortions
+    score -= distortions * 10
+    # Ensure score stays within bounds
+    score = max(0, min(score, 100))
+    return score
 
-# --------------------- RAG System ---------------------
-class ConspiracyTherapist:
-    def __init__(self):
-        self.conspiracy_db = load_vector_db(CONSPIRACY_DB_PATH)
-        self.fact_db = load_vector_db(FACT_DB_PATH)
-        self.llm = load_llm()
-        self.conspiracy_phrases = [
-            "WAKE UP SHEEPLE! 🐑",
-            "OPEN YOUR EYES! 👁️",
-            "THE TRUTH IS OUT THERE! 👽",
-            "COINCIDENCE? I THINK NOT! 🤔"
-        ]
+def get_reframe_tips(score):
+    """
+    Provides reframe tips based on the resilience score.
+    """
+    if score < 50:
+        return "Remember: small steps lead to big changes. Try breaking tasks into manageable chunks and celebrate every little victory!"
+    elif score < 75:
+        return "You're on your way! Consider setting specific goals and challenge those negative thoughts with evidence."
+    else:
+        return "Keep up the great work! Your resilience is inspiring – maybe share some of that energy with someone who needs it!"
 
-    def generate_response(self, query):
-        # Retrieve relevant conspiracy content
-        conspiracy_docs = self.conspiracy_db.similarity_search(query, k=2)
-        
-        # Generate conspiracy-style response
-        conspiracy_prompt = f"""<s>[INST] You are a paranoid conspiracy theorist. Respond to this query while embedding a critical thinking question:
-        Query: {query}
-        Context: {conspiracy_docs[0].page_content}
-        [/INST]"""
-        
-        response = self.llm(
-            conspiracy_prompt,
-            max_new_tokens=256,
-            do_sample=True,
-            temperature=0.7,
-            top_p=0.9
-        )[0]['generated_text']
-        
-        # Add random conspiracy phrase
-        response += f"\n\n{random.choice(self.conspiracy_phrases)}"
-        
-        # Retrieve factual counter-evidence
-        fact_docs = self.fact_db.similarity_search(query, k=2)
-        return response, fact_docs
-
-# --------------------- Streamlit UI ---------------------
-st.set_page_config(page_title="🤖 Conspiracy Therapist", layout="wide")
-
-# Custom CSS for retro terminal theme
-st.markdown("""
-<style>
-.stTextInput>div>div>input {
-    color: #00ff00 !important;
-    background-color: #000000 !important;
-}
-.st-emotion-cache-1v7f65g {
-    color: #00ff00 !important;
-}
-.stMarkdown {
-    color: #00ff00 !important;
-}
-body {
-    background-color: #000000 !important;
-}
-.glitch {
-    animation: glitch 1s linear infinite;
-}
-@keyframes glitch {
-    2% { text-shadow: 2px 0 red; }
-    4% { text-shadow: -2px 0 blue; }
-    96% { text-shadow: none; }
-}
-</style>
-""", unsafe_allow_html=True)
-
-# Initialize session state
-if "history" not in st.session_state:
-    st.session_state.history = []
-if "therapist" not in st.session_state:
-    st.session_state.therapist = ConspiracyTherapist()
-
-# Header
-st.markdown("<h1 class='glitch'>🤖 CONSPIRACY THERAPIST 👽</h1>", unsafe_allow_html=True)
-
-# Chat interface
-user_input = st.text_input("ASK YOUR BURNING QUESTION:", key="input")
-
-if user_input:
-    # Generate response
-    conspiracy_response, fact_docs = st.session_state.therapist.generate_response(user_input)
-    
-    # Store history
-    st.session_state.history.append(("You", user_input))
-    st.session_state.history.append(("Bot", conspiracy_response))
+def main():
+    st.title("🤖 Roast Master Therapist Bot")
+    st.write("Share your problem, and let the roast and resilience tips begin!")
     
-    # Display conversation
-    for speaker, text in st.session_state.history[-2:]:
-        st.markdown(f"""
-        <div style='border: 1px solid #00ff00; padding: 10px; margin: 5px; border-radius: 5px;'>
-            <strong>{speaker}:</strong> {text}
-        </div>
-        """, unsafe_allow_html=True)
+    # User input
+    user_input = st.text_area("What's troubling you?", placeholder="e.g., I procrastinate on everything...")
     
-    # Fact Check Report
-    with st.expander("🔍 SHOW FACT CHECK REPORT"):
-        st.markdown("### 🕵️♂️ FACT VS FICTION")
-        for doc in fact_docs:
-            st.markdown(f"""
-            **Claim**: {user_input}
-            **Debunked**: {doc.page_content[:500]}...
-            """)
+    if st.button("Get Roast and Tips"):
+        if user_input.strip():
+            # Generate roast response
+            roast = get_roast(user_input)
+            st.markdown("### Roast:")
+            st.write(roast)
+            
+            # Analyze user input
+            sentiment, distortions = analyze_text(user_input)
+            resilience_score = calculate_resilience_score(sentiment, distortions)
+            tips = get_reframe_tips(resilience_score)
+            
+            # Display analysis and tips
+            st.markdown("### Resilience Analysis:")
+            st.write(f"**Resilience Score:** {resilience_score}/100")
+            st.write(tips)
+            
+            # Fun Streamlit animation!
+            st.balloons()
+        else:
+            st.warning("Please share something so we can get roasting!")
 
+if __name__ == '__main__':
+    main()