🐛 Fix: improved model output post-processing to avoid empty answers

app.py

@@ -12,13 +12,14 @@ from langchain_community.embeddings import HuggingFaceEmbeddings
 from langchain.text_splitter import RecursiveCharacterTextSplitter
 from langchain_core.documents import Document
 
-# Load Reasoning Model (lightweight + CPU friendly)
-model_id = "google/flan-t5-base"
+#extra
+# Load Reasoning Model (CPU-friendly with better reasoning)
+model_id = "microsoft/phi-1_5"
 tokenizer = AutoTokenizer.from_pretrained(model_id)
-model = AutoModelForSeq2SeqLM.from_pretrained(model_id)
+model = AutoModelForCausalLM.from_pretrained(model_id)
 
 reasoning_pipeline = pipeline(
-    "text2text-generation",
+    "text-generation",
     model=model,
     tokenizer=tokenizer,
     max_new_tokens=512,
@@ -33,6 +34,7 @@ vectorstore = None
 # Summarizer
 summary_pipeline = pipeline("summarization", model="sshleifer/distilbart-cnn-12-6")
 
+
 def clean_text(text):
     lines = text.split("\n")
     cleaned = []