Project description
SIT Student Policy Advisor is a document question-answering app for student policies. A user uploads at least two official, text-based PDF documents and asks a question. The app finds relevant passages, then uses Google’s Gemma model to answer from those passages, citing the source document and page. It also displays the retrieved passages so users can check the evidence. If the passages seem insufficiently relevant, it declines to answer. README · App code
Technologies used

• Python — the app’s programming language.
• Gradio 6 — builds the web interface, including PDF upload, chat, evidence display, and the evaluation worksheet.
• pypdf — extracts text and page numbers from PDFs.
• Sentence Transformers with all-MiniLM-L6-v2 — turns document passages and questions into embeddings for semantic search.
• NumPy — compares embeddings to rank the most relevant passages.
• Google Gen AI SDK with Gemma 4 26B A4B Instruct by default — generates answers from the retrieved passages. A Google API key is needed for generated answers; without one, the app can still show retrieved passages.
• Jupyter/Google Colab — the repo also includes the adapted starter-kit notebook.
• pip / requirements.txt — installs the Python dependencies.