Update app.py

app.py

@@ -3,11 +3,15 @@ import pdfplumber
 import faiss
 import numpy as np
 from sentence_transformers import SentenceTransformer
-from transformers import pipeline
+from transformers import AutoModelForCausalLM, AutoTokenizer
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 
-# Load Models
-qa_pipeline = pipeline("question-answering", model="deepset/roberta-base-squad2")
+# Load Mistral-7B Model for Generative Answers
+model_name = "mistralai/Mistral-7B-Instruct"  # More advanced model
+tokenizer = AutoTokenizer.from_pretrained(model_name)
+model = AutoModelForCausalLM.from_pretrained(model_name)
+
+# Load Sentence Embeddings Model for Better Context Matching
 embedding_model = SentenceTransformer("all-MiniLM-L6-v2")
 
 # Function to Extract & Clean PDF Text
@@ -36,6 +40,13 @@ def find_best_chunk(question, index, chunks, embeddings):
     _, closest_idx = index.search(np.array(question_embedding), 1)
     return chunks[closest_idx[0][0]]
 
+# Function to Generate a Long, Detailed Answer
+def get_answer(question, context):
+    input_text = f"Question: {question}\nContext: {context}\nAnswer:"
+    inputs = tokenizer(input_text, return_tensors="pt")
+    output = model.generate(**inputs, max_length=300, temperature=0.7)
+    return tokenizer.decode(output[0], skip_special_tokens=True)
+
 # Streamlit UI
 st.title("Chat with AWS Restart PDF (Like ChatPDF)")
 
@@ -52,5 +63,5 @@ question = st.text_input("Ask a question about AWS Restart program:")
 
 if st.button("Get Answer") and question:
     relevant_chunk = find_best_chunk(question, index, chunks, embeddings)
-    response = qa_pipeline(question=question, context=relevant_chunk)
-    st.write("Answer:", response['answer'])
+    response = get_answer(question, relevant_chunk)
+    st.write("Answer:", response)