Create app.py

app.py

@@ -0,0 +1,72 @@
+import gradio as gr
+from transformers import pipeline
+import re
+
+sentiment_pipeline = pipeline("sentiment-analysis", model="distilbert-base-uncased-finetuned-sst-2-english")
+
+moderation_guidelines = """
+- Allow positive messages
+- Block cuss words
+- Allow negative comments about individuals but block negative comments against a community
+- Block words: Darren
+"""
+
+
+default_cuss_words = {
+    "damn", "hell", "shit", "fuck", "ass", "bastard", "bitch", "bollocks", "bugger", 
+    "bullshit", "crap", "dammit", "douche", "dumbass", "faggot", "jackass", "jerk", 
+    "motherfucker", "piss", "prick", "slut", "son of a bitch", "twat", "wanker"
+}
+
+community_terms = {"religion", "race", "ethnicity", "group", "community", "gender", "china"}
+
+def extract_blocked_words(guidelines):
+    """Extract blocked words from guidelines"""
+    match = re.search(r"block words:\s*(.*)", guidelines.lower())
+    return set(match.group(1).split(",")) if match else set()
+
+def moderate_message(message, guidelines):
+    """Moderates a message based on sentiment and dynamic guidelines."""
+    
+    sentiment = sentiment_pipeline(message)[0]['label']  
+    
+  
+    blocked_words = extract_blocked_words(guidelines)
+    
+    
+    allow_positive = "allow positive" in guidelines.lower()
+    block_cuss_words = "block cuss" in guidelines.lower()
+    allow_negative_personal = "allow negative comments about individuals" in guidelines.lower()
+    block_negative_community = "block negative comments against a community" in guidelines.lower()
+
+  
+    words = set(re.findall(r'\w+', message.lower()))  
+    if block_cuss_words and (words & default_cuss_words):
+        return "❌ Message Blocked: Contains inappropriate language."
+
+    if words & blocked_words:
+        return "🚫 Message Blocked: Contains restricted words."
+    if sentiment == "POSITIVE" and allow_positive:
+        return f"✅ Allowed (Positive): {message}"
+
+    if sentiment == "NEGATIVE":
+        if any(word in message.lower() for word in community_terms) and block_negative_community:
+            return "🚫 Message Blocked: Negative content targeting a community."
+        elif allow_negative_personal:
+            return f"⚠️ Allowed (Negative - Personal Attack): {message}"
+    
+    return f"✅ Allowed (Neutral): {message}"
+
+with gr.Blocks() as demo:
+    gr.Markdown("### 🛡️ AI-Powered Moderation System")
+
+    guidelines_input = gr.Textbox(value=moderation_guidelines, label="Moderation Guidelines (Admins Can Update)", lines=4)
+
+    with gr.Row():
+        msg_input = gr.Textbox(label="Enter Message")
+        msg_output = gr.Textbox(label="Moderation Result", interactive=False)
+    
+    moderate_btn = gr.Button("Check Message")
+    moderate_btn.click(moderate_message, inputs=[msg_input, guidelines_input], outputs=[msg_output])
+
+demo.launch()