Metric Recording Implementation

app.py

@@ -2,6 +2,7 @@ import gradio as gr
 from langchain.prompts import ChatPromptTemplate
 from langchain.schema import HumanMessage, SystemMessage, AIMessage
 from huggingface_hub import InferenceClient
+from metrics import EduBotMetrics
 import os
 import time
 import logging
@@ -21,6 +22,9 @@ client = InferenceClient(
     provider="together",
     api_key=hf_token,
 )
+
+metrics_tracker = EduBotMetrics()
+
 math_template = ChatPromptTemplate.from_messages([
     ("system", """{system_message}
 You are an expert math tutor. For every math problem:
@@ -103,6 +107,14 @@ def smart_truncate(text, max_length=3000):
 
 def respond_with_enhanced_streaming(message, history):
     """Streams the bot's response, detecting the subject and handling errors."""
+
+    # Start metrics timing
+    timing_context = metrics_tracker.start_timing()
+    error_occurred = False
+    error_message = None
+    response = ""
+    mode = ""
+    
     try:
         template, mode = detect_subject(message)
         
@@ -139,6 +151,9 @@ def respond_with_enhanced_streaming(message, history):
             {"role": "system", "content": "You are EduBot, an expert AI learning assistant."},
             {"role": "user", "content": formatted_prompt}
         ]
+
+        # Mark provider API start
+        metrics_tracker.mark_provider_start(timing_context)
         
         completion = client.chat.completions.create(
             model="Qwen/Qwen2.5-7B-Instruct",
@@ -147,9 +162,11 @@ def respond_with_enhanced_streaming(message, history):
             temperature=0.7,
             top_p=0.9,
         )
+
+        # Mark provider API end
+        metrics_tracker.mark_provider_end(timing_context)
         
         response = completion.choices[0].message.content
-        
         response = smart_truncate(response, max_length=3000)
         
         # Stream the response word by word
@@ -159,19 +176,36 @@ def respond_with_enhanced_streaming(message, history):
         for i, word in enumerate(words):
             partial_response += word + " "
             
-            # Update the stream periodically
+            # Update the stream periodically and record chunks
             if i % 4 == 0:
+                metrics_tracker.record_chunk(timing_context)
                 yield partial_response
                 time.sleep(0.03)
         
         final_response = f"*{mode}*\n\n{response}"
         logger.info(f"Response completed. Length: {len(response)} characters")
+        
+        # Record final chunk
+        metrics_tracker.record_chunk(timing_context)
         yield final_response
         
     except Exception as e:
+        error_occurred = True
+        error_message = str(e)
         logger.exception("Error in response generation")
         yield f"Sorry, I encountered an error: {str(e)}"
 
+    finally:
+        # Log the complete interaction with metrics
+        metrics_tracker.log_interaction(
+            mode=mode or "Unknown",
+            query=message,
+            response=response,
+            timing_context=timing_context,
+            error_occurred=error_occurred,
+            error_message=error_message
+        )
+
 # --- Fixed Gradio UI and CSS ---
 custom_css = """
 /* Main container styling */