Update app/main.py

app/main.py

@@ -9,10 +9,10 @@ from langchain.chains import create_retrieval_chain
 from langchain_community.vectorstores import FAISS
 from langchain_community.document_loaders import UnstructuredWordDocumentLoader as DocxLoader
 from fastapi.middleware.cors import CORSMiddleware
-from fastapi import FastAPI, Request
+from fastapi import FastAPI
 from pydantic import BaseModel
 from langchain_community.embeddings import HuggingFaceBgeEmbeddings
-import nltk
+import nltk  # Importing NLTK
 import time
 
 # Set writable paths for cache and data
@@ -43,10 +43,18 @@ except Exception as e:
     raise
 
 def clean_response(response):
+    # Remove any leading/trailing whitespace, including newlines
     cleaned = response.strip()
-    cleaned = re.sub(r'^\"|\"$', '', cleaned)
+    
+    # Remove any enclosing quotation marks
+    cleaned = re.sub(r'^["\']+|["\']+$', '', cleaned)
+    
+    # Replace multiple newlines with a single newline
     cleaned = re.sub(r'\n+', '\n', cleaned)
+    
+    # Remove any remaining '\n' characters
     cleaned = cleaned.replace('\\n', '')
+    
     return cleaned
 
 app = FastAPI()
@@ -62,32 +70,29 @@ app.add_middleware(
 openai_api_key = os.environ.get('OPENAI_API_KEY')
 llm = ChatOpenAI(
     api_key=openai_api_key,
-    model_name="gpt-4-turbo-preview",
+    model_name="gpt-4-turbo-preview",  # or "gpt-3.5-turbo" for a more economical option
     temperature=0.7
 )
 
-conversation_history = {}  # Dictionary to maintain contextual memory
-
 @app.get("/")
 def read_root():
     return {"Hello": "World"}
 
 class Query(BaseModel):
-    session_id: str  # Unique identifier for user session
     query_text: str
 
-prompt_template = ChatPromptTemplate.from_template(
-    """
-    You are a helpful assistant designed specifically for the Thapar Institute of Engineering and Technology (TIET), a renowned technical college. Your task is to answer all queries related to TIET. Every response you provide should be relevant to the context of TIET. If a question falls outside of this context, please decline by stating, 'Sorry, I cannot help with that.'
-
-    If the query is not related to TIET or falls outside the context of education, respond with:
-    "Sorry, I cannot help with that. I'm specifically designed to answer questions about the Thapar Institute of Engineering and Technology. For more information, please contact at our toll-free number: 18002024100 or E-mail us at admissions@thapar.edu"
-
-    <context>
-    {context}
-    </context>
-    Question: {input}
-    """
+prompt = ChatPromptTemplate.from_template(
+"""
+You are a helpful assistant designed specifically for the Thapar Institute of Engineering and Technology (TIET), a renowned technical college. Your task is to answer all queries related to TIET. Every response you provide should be relevant to the context of TIET. If a question falls outside of this context, please decline by stating, 'Sorry, I cannot help with that.' If you do not know the answer to a question, do not attempt to fabricate a response; instead, politely decline.
+You may elaborate on your answers slightly to provide more information, but avoid sounding boastful or exaggerating. Stay focused on the context provided.
+If the query is not related to TIET or falls outside the context of education, respond with:
+        "Sorry, I cannot help with that. I'm specifically designed to answer questions about the Thapar Institute of Engineering and Technology. 
+        For more information, please contact at our toll-free number: 18002024100 or E-mail us at admissions@thapar.edu
+<context>
+{context}
+</context>
+Question: {input}  
+"""
 )
 
 def vector_embedding():
@@ -104,16 +109,16 @@ def vector_embedding():
 
         text_splitter = RecursiveCharacterTextSplitter(chunk_size=500, chunk_overlap=100)
         chunks = text_splitter.split_documents(documents)
-
+        
         print(f"Created {len(chunks)} chunks.")
 
         model_name = "BAAI/bge-base-en"
         encode_kwargs = {'normalize_embeddings': True}
         model_norm = HuggingFaceBgeEmbeddings(model_name=model_name, encode_kwargs=encode_kwargs)
-
+        
         db = FAISS.from_documents(chunks, model_norm)
         db.save_local("./vectors_db")
-
+        
         print("Vector store created and saved successfully.")
         return {"response": "Vector Store DB Is Ready"}
 
@@ -127,40 +132,33 @@ def get_embeddings():
     model_norm = HuggingFaceBgeEmbeddings(model_name=model_name, encode_kwargs=encode_kwargs)
     return model_norm
 
-@app.post("/chat")
-def chat_endpoint(query: Query):
+@app.post("/chat")  # Changed from /anthropic to /chat
+def read_item(query: Query):
     try:
-        session_id = query.session_id
-        if session_id not in conversation_history:
-            conversation_history[session_id] = []
-
         embeddings = get_embeddings()
         vectors = FAISS.load_local("./vectors_db", embeddings, allow_dangerous_deserialization=True)
     except Exception as e:
         print(f"Error loading vector store: {str(e)}")
         return {"response": "Vector Store Not Found or Error Loading. Please run /setup first."}
-
+    
     prompt1 = query.query_text
     if prompt1:
         start = time.process_time()
-        document_chain = create_stuff_documents_chain(llm, prompt_template)
+        document_chain = create_stuff_documents_chain(llm, prompt)
         retriever = vectors.as_retriever()
         retrieval_chain = create_retrieval_chain(retriever, document_chain)
-
-        # Combine context from conversation history
-        context = "\n".join(conversation_history[session_id])
-        response = retrieval_chain.invoke({'input': prompt1, 'context': context})
-
-        cleaned_response = clean_response(response['answer'])
-
-        # Update conversation history
-        conversation_history[session_id].append(f"User: {prompt1}")
-        conversation_history[session_id].append(f"Assistant: {cleaned_response}")
-
+        response = retrieval_chain.invoke({'input': prompt1})
         print("Response time:", time.process_time() - start)
-        return {"response": cleaned_response}
+        
+        # Apply the cleaning function to the response
+        cleaned_response = clean_response(response['answer'])
+        
+        # For debugging, print the cleaned response
+        print("Cleaned response:", repr(cleaned_response))
+        
+        return cleaned_response
     else:
-        return {"response": "No Query Found"}
+        return "No Query Found"
 
 @app.get("/setup")
 def setup():
@@ -168,4 +166,4 @@ def setup():
 
 if __name__ == "__main__":
     import uvicorn
-    uvicorn.run(app, host="0.0.0.0", port=8000)
+    uvicorn.run(app, host="0.0.0.0", port=8000)