Update app.py

app.py

@@ -4,6 +4,7 @@ import gradio as gr
 from transformers import pipeline
 from sentence_transformers import SentenceTransformer, util
 import PyPDF2
+import re
 
 # Set up logging with immediate writing
 logging.basicConfig(
@@ -18,6 +19,43 @@ logger = logging.getLogger()
 qa_model = pipeline("question-answering", model="distilbert-base-uncased-distilled-squad")
 embedder = SentenceTransformer('all-MiniLM-L6-v2')
 
+import re
+
+def preprocess_qa_format(text):
+    """
+    Detects Q/A pairs in different formats and merges them so that the answer appears first.
+
+    Supported formats:
+    1. "question 69: what is xyz?\nanswer:: dark chocolate"
+    2. "Q 21: what is xyz?\nA: It is a new form of tyre."
+    3. "Q 21: what is xyz?\nit is a new form of tyre."
+    
+    Transforms them into:
+    "Answer: [answer]\nQuestion: [question]"
+    """
+
+    # Matches explicit "Question X: ..." followed by "Answer: ..."
+    pattern1 = re.compile(r"(?i)question\s*\d*\s*:\s*(.+?)\n\s*answer[:]*\s*(.+?)(?:\n|$)")
+    
+    # Matches "Q X: ..." followed by "A: ..."
+    pattern2 = re.compile(r"(?i)Q\s*\d*\s*:\s*(.+?)\n\s*A[:]*\s*(.+?)(?:\n|$)")
+
+    # Matches "Q X: ..." followed by a sentence as an inferred answer (not explicitly labeled)
+    pattern3 = re.compile(r"(?i)Q\s*\d*\s*:\s*(.+?)\n\s*([A-Z][^.]*\..+?)(?:\n|$)")
+
+    def replacer(match):
+        question_text = match.group(1).strip()
+        answer_text = match.group(2).strip()
+        return f"Answer: {answer_text}\nQuestion: {question_text}\n"
+
+    # Apply transformations
+    text = pattern1.sub(replacer, text)
+    text = pattern2.sub(replacer, text)
+    text = pattern3.sub(replacer, text)
+
+    return text
+
+
 # Helper function to extract text from PDF
 def extract_text_from_pdf(file_path):
     text = ""