Update app.py

app.py

@@ -1,22 +1,44 @@
 import streamlit as st
-from transformers import pipeline
+from transformers import AutoTokenizer, AutoModelForCausalLM
+import torch
 
-# Load Hugging Face classifier
-classifier = pipeline("zero-shot-classification")
+st.title("🚨 Fake Job Posting Detector")
+st.write(
+    "Enter a job description below to check if it is likely **Fake** or **Legit**. "
+    "This tool uses AI to help job seekers avoid scams."
+)
 
-st.title("Fake Job Posting Detector")
+# Load the model
+@st.cache_resource(show_spinner=True)
+def load_model():
+    model_id = "openai/gpt-oss-20b"
+    tokenizer = AutoTokenizer.from_pretrained(model_id)
+    model = AutoModelForCausalLM.from_pretrained(model_id)
+    return tokenizer, model
 
-# Input job description
-job_description = st.text_area("Enter the job description:")
+tokenizer, model = load_model()
 
-if st.button("Check Job"):
+# Input text
+job_description = st.text_area(
+    "Paste the job description here:",
+    "Example: Urgent hiring! Work from home, no experience needed, $5000/month!"
+)
+
+# Button to run prediction
+if st.button("Check Job Posting"):
     if job_description.strip() == "":
-        st.warning("Please enter a job description.")
+        st.warning("⚠️ Please enter a job description first.")
     else:
-        labels = ["Legit", "Fake"]
-        result = classifier(job_description, candidate_labels=labels)
-        predicted_label = result['labels'][0]
-        confidence = result['scores'][0]
-        
-        st.write(f"Prediction: **{predicted_label}**")
-        st.write(f"Confidence: {confidence:.2f}")
+        # Prepare prompt for GPT-OSS
+        prompt = f"Classify the following job posting as Fake or Legit:\n\n{job_description}\n\nAnswer with only 'Fake' or 'Legit'."
+
+        inputs = tokenizer(prompt, return_tensors="pt")
+        with torch.no_grad():
+            outputs = model.generate(**inputs, max_new_tokens=20)
+        prediction = tokenizer.decode(outputs[0], skip_special_tokens=True)
+
+        # Display result with color
+        if "Fake" in prediction:
+            st.error(f"Prediction: Fake")
+        else:
+            st.success(f"Prediction: Legit")