Lab: The Tidewell inbox service, end to end

Module 11 · Capstone lab · About 120 minutes · Needs: the labs of Modules 1 to 10 (or their cells, pasted here). CPU only; a Groq or Gemini key only for the LLM track

What you are building

It is Monday 16 November 2026, project week 11, and Rosa signs off the service today. Ten labs left ten pieces in tidewell-inbox/. This lab joins them into process(message, parts): one message in, one record out, all five jobs done. For each job it keeps the cheapest method that meets Rosa's bar, chosen from three build tracks: classical, transformer and LLM, the last switched off and checked with scripted stand-ins. The evidence is shared: per-language scores on HARD_TEST, a cost and latency table measured on this CPU, and the name-swap check.

The hand-over follows a published idea. Margaret Mitchell and colleagues, in "Model Cards for Model Reporting" (FAT* 2019, arXiv:1810.03993), proposed shipping a model with a short document: what it is for and not for, how it was evaluated, its results by group and condition, and its known limits. Your MODEL_CARD.md follows that idea.

Open a new Colab notebook, call it module-11-capstone.ipynb, and run each cell in order.