Update app.py

app.py

@@ -1,50 +1,87 @@
 import gradio as gr
-from transformers import AutoModelForCausalLM, AutoTokenizer
+import numpy as np
+import pandas as pd
+from transformers import AutoTokenizer, AutoModelForCausalLM
 import torch
 
-MODEL_NAME = "nvidia/OpenGPT-OSS-20B"
-
-print("Loading model... this may take some time.")
+MODEL_NAME = "microsoft/phi-2"
 
 tokenizer = AutoTokenizer.from_pretrained(MODEL_NAME)
 model = AutoModelForCausalLM.from_pretrained(
     MODEL_NAME,
-    torch_dtype=torch.float16 if torch.cuda.is_available() else torch.float32,
-    device_map="auto"
+    torch_dtype=torch.float32,
+    device_map="cpu"
 )
 
-def chat(prompt, history):
-    messages = ""
-    for user, bot in history:
-        messages += f"User: {user}\nAssistant: {bot}\n"
-    messages += f"User: {prompt}\nAssistant:"
-
-    inputs = tokenizer(messages, return_tensors="pt").to(model.device)
-    outputs = model.generate(
-        **inputs,
-        max_new_tokens=300,
-        temperature=0.7,
-        do_sample=True,
-        top_p=0.9,
-    )
-    response = tokenizer.decode(outputs[0], skip_special_tokens=True)
-    # Remove the previous conversation from output
-    reply = response.split("Assistant:")[-1].strip()
-    history.append((prompt, reply))
-    return reply, history
+# ----- Trend detection (math-based, NOT LLM) -----
+def detect_trend(values):
+    diffs = np.diff(values)
+
+    if all(d > 0 for d in diffs):
+        return "INCREASING"
+    elif all(d < 0 for d in diffs):
+        return "DECREASING"
+    else:
+        return "MIXED"
+
+# ----- Simple anomaly detection -----
+def detect_anomaly(values):
+    mean = np.mean(values)
+    std = np.std(values)
+
+    anomalies = []
+    for i, v in enumerate(values):
+        if abs(v - mean) > 2 * std:
+            anomalies.append((i, v))
+
+    if len(anomalies) == 0:
+        return "No anomalies detected"
+    return str(anomalies)
+
+# ----- LLM explanation -----
+def explanation(entity, values, trend):
+    prompt = f"""
+You are a KPI analysis expert.
+The entity is: {entity}
+The values are: {values}
+The detected trend is: {trend}
 
+Explain in simple words why the trend is {trend}.
+"""
+    inputs = tokenizer(prompt, return_tensors="pt")
+    outputs = model.generate(**inputs, max_new_tokens=150)
+    return tokenizer.decode(outputs[0], skip_special_tokens=True)
 
+# ----- Main function -----
+def analyze(entity, value_string):
+    try:
+        values = [float(x.strip()) for x in value_string.split(",")]
+    except:
+        return "❌ Error: Please enter numbers separated by commas"
+
+    trend = detect_trend(values)
+    anomaly = detect_anomaly(values)
+    exp = explanation(entity, values, trend)
+
+    return f"""
+📌 **Entity:** {entity}
+📉 **Trend:** {trend}
+⚠️ **Anomalies:** {anomaly}
+
+🧠 **Explanation:**  
+{exp}
+"""
+
+# ----- UI -----
 with gr.Blocks() as demo:
-    gr.Markdown("# 🧠 GPT-OSS-20B Chat (HuggingFace Space)")
-    chatbot = gr.Chatbot()
-    message = gr.Textbox(label="Type your message")
-    clear = gr.Button("Clear Chat")
+    gr.Markdown("# 📈 KPI Trend + Anomaly Analyzer (CPU-Friendly)")
+
+    entity = gr.Textbox(label="KPI Name (example: volte_rate)")
+    values = gr.Textbox(label="Values (comma separated, example: 10,11,13,14,15)")
+    output = gr.Markdown()
 
-    def user_submit(msg, history):
-        reply, history = chat(msg, history)
-        return "", history
+    run_btn = gr.Button("Analyze")
 
-    message.submit(user_submit, [message, chatbot], [message, chatbot])
-    clear.click(lambda: None, None, chatbot)
+    run_btn.click(fn=analyze, inputs=[entity, values], outputs=output)
 
 demo.launch()