Remove source URL from rejection responses

Show source URL on rejection responses & update notebook outputs

Make embeddings cache loading robust with try/except

Fix numpy allow_pickle error on cached embeddings

Always include source URL; reject results without one

Fix AttributeError on NaN url values

Switch LLM backend from Ollama to Groq

Fix 'Cannot assign requested address' error and add LLM fallback

Add pre-computed embeddings cache and optimize load_pipeline

Replace broken Excel-as-CSV with properly generated UTF-8 CSV

Read Excel dataset directly instead of broken CSV

Use latin-1 encoding to read CSV, export CSV as UTF-8

Fix UnicodeDecodeError when reading MedQuAD CSV

Add Streamlit chat UI (app.py) for deployable medical chatbot