Update app.py

app.py

@@ -2,15 +2,17 @@ import gradio as gr
 from huggingface_hub import InferenceClient
 from typing import List, Tuple
 import logging
+from collections import deque
+import re
 
-# Configure logging for better debugging and monitoring
+# Configure logging
 logging.basicConfig(
     level=logging.INFO,
     format="%(asctime)s - %(levelname)s - %(message)s",
 )
 logger = logging.getLogger(__name__)
 
-# Initialize the InferenceClient with error handling
+# Initialize the InferenceClient
 try:
     client = InferenceClient("HuggingFaceH4/zephyr-7b-beta")
     logger.info("Successfully initialized InferenceClient")
@@ -18,6 +20,34 @@ except Exception as e:
     logger.error(f"Failed to initialize InferenceClient: {str(e)}")
     raise
 
+# Memory storage for learning from past queries
+MEMORY = deque(maxlen=100)  # Store up to 100 query-response pairs
+
+def add_to_memory(query: str, response: str):
+    """Add a query-response pair to memory."""
+    MEMORY.append({"query": query, "response": response})
+    logger.info("Added query-response pair to memory")
+
+def find_relevant_context(query: str, max_contexts: int = 2) -> str:
+    """Retrieve relevant past queries and responses based on simple keyword matching."""
+    query_words = set(re.findall(r'\w+', query.lower()))
+    relevant = []
+    
+    for mem in MEMORY:
+        mem_words = set(re.findall(r'\w+', mem["query"].lower()))
+        overlap = len(query_words & mem_words) / max(len(query_words), 1)
+        if overlap > 0.3:  # Threshold for relevance
+            relevant.append(mem)
+        if len(relevant) >= max_contexts:
+            break
+    
+    if relevant:
+        context = "\n".join(
+            [f"Past Query: {mem['query']}\nPast Response: {mem['response']}" for mem in relevant]
+        )
+        return f"Relevant past interactions:\n{context}\n\n"
+    return ""
+
 def respond(
     message: str,
     history: List[Tuple[str, str]],
@@ -27,7 +57,7 @@ def respond(
     top_p: float,
 ) -> str:
     """
-    Generates an educational response to a student's query using the HuggingFace Inference API.
+    Generates an educational response using past interactions for context.
     Args:
         message (str): The student's input question or query.
         history (List[Tuple[str, str]]): Chat history with student and AI teacher messages.
@@ -37,9 +67,6 @@ def respond(
         top_p (float): Controls diversity via nucleus sampling.
     Yields:
         str: The AI teacher's response, streamed token by token.
-    Raises:
-        ValueError: If input parameters are invalid.
-        RuntimeError: If the API call fails.
     """
     # Validate input parameters
     if not message.strip():
@@ -51,8 +78,16 @@ def respond(
     if top_p < 0.1 or top_p > 1.0:
         raise ValueError("top_p must be between 0.1 and 1.0")
 
-    # Construct the message history
-    messages = [{"role": "system", "content": system_message}]
+    # Retrieve relevant past interactions
+    context = find_relevant_context(message)
+    
+    # Construct the message history with memory context
+    messages = [
+        {
+            "role": "system",
+            "content": system_message + "\n\nUse the following past interactions to inform your response if relevant:\n" + context,
+        }
+    ]
     for user_msg, assistant_msg in history:
         if user_msg:
             messages.append({"role": "user", "content": user_msg})
@@ -72,6 +107,8 @@ def respond(
             token = message.choices[0].delta.content or ""
             response += token
             yield response
+        # Store the query and final response in memory
+        add_to_memory(message, response)
     except Exception as e:
         logger.error(f"Error during chat completion: {str(e)}")
         raise RuntimeError("Failed to generate response from the model")
@@ -80,7 +117,6 @@ def main():
     """
     Sets up and launches the Gradio ChatInterface for the AI Teacher chatbot.
     """
-    # Define default system message for an AI teacher
     default_system_message = (
         "You are an AI Teacher, a knowledgeable and patient educator dedicated to helping students and learners. "
         "Your goal is to explain concepts clearly, provide step-by-step guidance, and encourage critical thinking. "
@@ -88,7 +124,6 @@ def main():
         "Be supportive, professional, and engaging in all interactions."
     )
 
-    # Create Gradio ChatInterface with settings compatible with older Gradio versions
     demo = gr.ChatInterface(
         fn=respond,
         additional_inputs=[
@@ -104,7 +139,6 @@ def main():
                 value=512,
                 step=1,
                 label="Maximum Response Length",
-                info="Controls the maximum length of the AI Teacher's response.",
             ),
             gr.Slider(
                 minimum=0.1,
@@ -112,7 +146,6 @@ def main():
                 value=0.7,
                 step=0.1,
                 label="Response Creativity",
-                info="Lower values make responses more focused and precise.",
             ),
             gr.Slider(
                 minimum=0.1,
@@ -120,7 +153,6 @@ def main():
                 value=0.95,
                 step=0.05,
                 label="Response Diversity",
-                info="Lower values focus on more likely and relevant answers.",
             ),
         ],
         title="AI Teacher: Your Study Companion",
@@ -138,7 +170,6 @@ def main():
         """,
     )
 
-    # Launch the application
     try:
         logger.info("Launching Gradio interface for AI Teacher")
         demo.launch(server_name="0.0.0.0", server_port=7860)