Update app.py

app.py

@@ -5,11 +5,9 @@ import fitz  # PyMuPDF
 from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain_community.vectorstores import Chroma
 from langchain.text_splitter import RecursiveCharacterTextSplitter
-from langchain_huggingface import HuggingFacePipeline
 from langchain.prompts import PromptTemplate
 from langchain.chains import LLMChain
-from transformers import pipeline
-import torch
+import anthropic
 import base64
 from PIL import Image
 import io
@@ -71,7 +69,7 @@ class CurriculumChatbot:
         )
     
     def _setup_llm(self):
-        """Setup LLM with HuggingFace pipeline"""
+        """Setup LLM with Claude"""
         try:
             # Initialize LLM attributes
             self.llm = None
@@ -79,20 +77,30 @@ class CurriculumChatbot:
             self.focused_qa_chain = None
             self.content_selection_chain = None
             
-            # Load the model
-            pipe = pipeline(
-                "text-generation",
-                model="microsoft/DialoGPT-medium",
-                torch_dtype=torch.float16,
-                device_map="auto",
-                max_length=512,
-                do_sample=True,
-                temperature=0.7,
-                top_p=0.9,
-                repetition_penalty=1.1
+            # Load Claude
+            self.anthropic_client = anthropic.Anthropic(
+                api_key=os.environ.get("ANTHROPIC_API_KEY")
             )
             
-            self.llm = HuggingFacePipeline(pipeline=pipe)
+            # Create a custom LLM wrapper that works with LangChain
+            class ClaudeLLM:
+                def __init__(self, client):
+                    self.client = client
+                
+                def __call__(self, prompt):
+                    try:
+                        response = self.client.messages.create(
+                            model="claude-3-5-haiku-20241022",
+                            max_tokens=1500,
+                            temperature=0.7,
+                            messages=[{"role": "user", "content": prompt}]
+                        )
+                        return response.content[0].text
+                    except Exception as e:
+                        print(f"Error calling Claude: {e}")
+                        return "I'm sorry, I couldn't generate a response at the moment."
+            
+            self.llm = ClaudeLLM(self.anthropic_client)
             
             # Create content selection prompt template
             content_selection_template = """You are an expert curriculum analyst. Your task is to find the most relevant slide for a student's question.
@@ -118,7 +126,7 @@ Most relevant slide number:"""
                 template=content_selection_template
             ))
             
-            # Create QA prompt template for DialoGPT
+            # Create QA prompt template for Claude
             qa_template = """You are an expert programming tutor. Your task is to provide a comprehensive, educational answer based on the curriculum content.
 
 Curriculum Content:
@@ -133,6 +141,7 @@ Instructions:
 4. Explain the concept step-by-step
 5. Make sure your answer directly addresses what the student is asking
 6. If the content is limited, provide additional educational context
+7. Structure your answer clearly with bullet points or numbered lists when appropriate
 
 Your detailed answer:"""
             
@@ -156,6 +165,7 @@ Instructions:
 4. Explain the concept step-by-step
 5. Make sure your answer directly addresses what the student is asking
 6. If the slide content is limited, provide additional educational context
+7. Structure your answer clearly with bullet points or numbered lists when appropriate
 
 Your detailed answer:"""
             
@@ -301,16 +311,14 @@ Your detailed answer:"""
                 
                 print(f"LLM Response: {answer[:200]}...")
                 
-                # Clean up the answer
+                # Clean up the answer (Claude is cleaner, but just in case)
                 answer = answer.strip()
-                if "<|eot_id|>" in answer:
-                    answer = answer.split("<|eot_id|>")[-1].strip()
                 
                 # Remove any prompt artifacts
                 if answer.startswith("Answer:"):
                     answer = answer[7:].strip()
-                if answer.startswith("Provide a clear, educational answer based on this slide:"):
-                    answer = answer[58:].strip()
+                if answer.startswith("Your detailed answer:"):
+                    answer = answer[20:].strip()
                 
                 # Check if the answer is too short, generic, or poor quality
                 if (len(answer.strip()) < 100 or 
@@ -345,10 +353,8 @@ Your detailed answer:"""
                 
                 answer = self.qa_chain.run(question=query, filled_context=filled_context)
                 answer = answer.strip()
-                if "<|eot_id|>" in answer:
-                    answer = answer.split("<|eot_id|>")[-1].strip()
-
-
+                
+                # Remove any prompt artifacts (Claude is cleaner, but just in case)
                 if answer.startswith("Answer:"):
                     answer = answer[7:].strip()
                 if answer.startswith("Provide a clear, educational answer explaining the concept:"):