edited again where it will just paste the output

chatbot.py

@@ -2,12 +2,10 @@ from transformers import T5Tokenizer, T5ForConditionalGeneration
 from sentence_transformers import SentenceTransformer
 from pinecone import Pinecone
 
-device = 'cpu' 
+device = 'cpu'
 
-# Initialize Pinecone instance
 pc = Pinecone(api_key='89eeb534-da10-4068-92f7-12eddeabe1e5')
 
-# Check if the index exists; if not, create it
 index_name = 'abstractive-question-answering'
 index = pc.Index(index_name)
 
@@ -23,14 +21,11 @@ def load_models():
 retriever, generator, tokenizer = load_models()
 
 def process_query(query):    
-    # Query Pinecone
     xq = retriever.encode([query]).tolist()
     xc = index.query(vector=xq, top_k=1, include_metadata=True)
     
-    # Print the response to check the structure
     print("Pinecone response:", xc)
 
-    # Check if 'matches' exists and is a list
     if 'matches' in xc and isinstance(xc['matches'], list):
         context = [m['metadata']['Output'] for m in xc['matches']]
         context_str = " ".join(context)
@@ -47,11 +42,10 @@ def process_query(query):
     inputs = tokenizer.encode(formatted_query, return_tensors="pt", max_length=512, truncation=True).to(device)
     ids = generator.generate(inputs, num_beams=2, min_length=10, max_length=60, repetition_penalty=1.2)
     answer = tokenizer.decode(ids[0], skip_special_tokens=True, clean_up_tokenization_spaces=False)
-
-    nli_keywords = ['not_equivalent', 'not_entailment', 'entailment', 'neutral']
     
-    # If any of the keywords are found in the answer, return the fallback message
+    nli_keywords = ['not_equivalent', 'not_entailment', 'entailment', 'neutral', 'not_enquiry']
+    
     if any(keyword in answer.lower() for keyword in nli_keywords):
-        return "Klasmeyt, can you elaborate your question?"
+        return context_str  
 
     return answer