fix: overhaul follow-up detection for conversation memory + add diagnostic logs

fix: switch HF endpoint to universal router URL (auto-selects provider)

fix: add stream error handling + startup diagnostics, bump CACHEBUST

fix: bump CACHEBUST to force Docker rebuild with updated model config

fix(backend): bust docker cache + fix async streaming generator to not block event loop

fix: clone dataset before creating subdirectories

fix: use git clone for dataset instead of hf-cli to bypass rate limits

feat: download 4.4GB data from dataset on build

feat: add Dockerfile and requirements.txt for HuggingFace deployment