Update app.py

app.py

@@ -6,6 +6,7 @@ import traceback
 import os
 import torch
 import numpy as np
+import pandas as pd  # <-- ADDED
 from datetime import datetime
 
 # ARF components
@@ -81,7 +82,7 @@ def generate_with_logprobs(prompt, max_new_tokens=100):
 # ----------------------------------------------------------------------
 # NLI detector
 # ----------------------------------------------------------------------
-nli_detector = NLIDetector()
+nli_detector = NLIDetector()  # Note: this file will be updated separately
 
 # ----------------------------------------------------------------------
 # Retrieval (sentence‑transformers + ChromaDB)
@@ -136,7 +137,6 @@ ai_risk_engine = AIRiskEngine()
 # IoT simulator
 # ----------------------------------------------------------------------
 iot_sim = IoTSimulator()
-iot_history = []  # store recent readings for prediction
 
 # ----------------------------------------------------------------------
 # Helper: update risk with feedback (global state – shared across users)
@@ -292,15 +292,16 @@ async def handle_audio(audio_file):
         logger.error(f"Audio task error: {e}")
         return {"error": str(e)}
 
-async def read_iot_sensors(fault_type):
+# MODIFIED: accept session state, return updated state and DataFrame
+async def read_iot_sensors(fault_type, history_state):
     """Read simulated IoT sensors, run diagnostics, predict failure, and return updated plot data."""
-    global last_task_category, iot_history
+    global last_task_category
     last_task_category = "iot"
     iot_sim.set_fault(fault_type if fault_type != "none" else None)
     data = iot_sim.read()
-    iot_history.append(data)
-    if len(iot_history) > 100:
-        iot_history.pop(0)
+    history_state.append(data)
+    if len(history_state) > 100:
+        history_state.pop(0)
 
     # Create IoTEvent with valid component name
     event = IoTEvent(
@@ -321,8 +322,8 @@ async def read_iot_sensors(fault_type):
 
     # Simple failure prediction
     prediction = None
-    if len(iot_history) >= 5:
-        temps = [h['temperature'] for h in iot_history[-5:]]
+    if len(history_state) >= 5:
+        temps = [h['temperature'] for h in history_state[-5:]]
         x = np.arange(len(temps))
         slope, intercept = np.polyfit(x, temps, 1)
         next_temp = slope * len(temps) + intercept
@@ -333,9 +334,14 @@ async def read_iot_sensors(fault_type):
                 "time_to_overheat_min": time_to_threshold
             }
 
-    # Prepare temperature history for plotting
-    temp_history = [h['temperature'] for h in iot_history[-20:]]  # last 20 readings
-    return data, diag_result, prediction, temp_history
+    # Prepare temperature history for plotting as DataFrame
+    temp_history = [h['temperature'] for h in history_state[-20:]]  # last 20 readings
+    df = pd.DataFrame({
+        "index": list(range(len(temp_history))),
+        "temperature": temp_history
+    })
+
+    return data, diag_result, prediction, df, history_state  # return updated state
 
 # ----------------------------------------------------------------------
 # Gradio UI
@@ -370,6 +376,8 @@ with gr.Blocks(title="ARF v4 – AI Reliability Lab", theme="soft") as demo:
         # Tab 4: Robotics / IoT (enhanced with live plot)
         with gr.TabItem("Robotics / IoT"):
             gr.Markdown("### Simulated Robotic Arm Monitoring")
+            iot_state = gr.State(value=[])  # per-session history state
+
             with gr.Row():
                 with gr.Column():
                     fault_type = gr.Dropdown(
@@ -438,10 +446,11 @@ with gr.Blocks(title="ARF v4 – AI Reliability Lab", theme="soft") as demo:
         inputs=audio_input,
         outputs=audio_output
     )
+    # MODIFIED: include state as input and output
     refresh_btn.click(
-        fn=lambda f: asyncio.run(read_iot_sensors(f)),
-        inputs=fault_type,
-        outputs=[sensor_display, diag_display, pred_display, temp_plot]
+        fn=lambda f, h: asyncio.run(read_iot_sensors(f, h)),
+        inputs=[fault_type, iot_state],
+        outputs=[sensor_display, diag_display, pred_display, temp_plot, iot_state]
     )
     feedback_up.click(fn=lambda: feedback(True), outputs=feedback_msg)
     feedback_down.click(fn=lambda: feedback(False), outputs=feedback_msg)