A recent study by the International Agency for Research on Cancer (IARC) reports a rising global burden of breast cancer, with an estimated one in twenty women expected to be diagnosed during their lifetime. The study, published in Nature Medicine, underscores a particularly concerning increase in incidence rates among countries with a low Human Development Index (HDI). In the existing research papers, most of the work has been done to categorize benign and malignant types but not their sub-types. The proposed work aims to enhance the classification accuracy of benign & malignant types and their sub types. A deep learning based Densenet121 model is developed to improve breast cancer sub-type classification accuracy, achieving 98.66% for malignant & benign, 94.35% for benign sub type classification and 92.12% for malignant sub-type classification. This work has been done on BreakHis dataset. Further for real time medical report generation, a secured diagnostic system has been developed that integrates Telegram bot interface with Flask API deployed on Raspberry Pi. Telegram bot is a breast cancer prediction assistant chatbot designed to predict breast tumor type based on histopathological image uploaded. The chatbot operates in three main modes: registration, consultation, and report generation. This system allows medical professionals to receive early diagnosis of cancer stage report. Early detection often requires less invasive treatments such as surgery or localized therapy instead of aggressive chemotherapy or radiotherapy.