Implemented exact Decimal totals, filtering, validation, and final ROUND_HALF_UP rounding. Added CLI `--provider` and concise error handling.

`python3 -m unittest -v`: all 10 tests passed. README.md and supplied tests unchanged.

Next prompt (Optional): "Review the implementation for edge cases."
Model: gpt-6-sol | Effort: low