Frame the grounded café FAQ application experiment

This lab builds grounded café FAQ app with validation, five-note retrieval, citations, tests, demo. The human in the loop is café ops lead answering hours/wifi/pets questions from a kiosk. Scope is intentionally narrower than “make AI reliable”: you will prove one oracle — gold 5/5 cited + 3/3 abstain; demo under 120s from clean checkout — and one invariant — unsupported questions abstain; every answer cites a note id; two-minute clean-room demo. Record non-goals in your notes so a later change cannot silently expand authority. The incident mnemonic for the chapter is CAP-CAFE-DEMO-01; design as if that ticket is already written and you are filling evidence.

Turn the oracle into a table: input fixture, expected observable, prohibited side effect, owner, latency/cost ceiling. Separate model taste from software correctness — transport, auth, parsing, and termination must be deterministic even when generated text varies. Primary metric family: gold_pass and demo_seconds. Averages without a denominator or revision label do not gate release. Fake external dependencies in unit tests; live calls wait until fakes pass.

Before implementation, encode a failing check that would have caught demo relies on personal .env and unreproducible notebook state. That failure is the pedagogical north star for later pages: contracts reject it, happy path never performs it, validation asserts it, failure-handling contains it, observability detects it, security-ops prevents privilege tricks around it, and mastery replays it in a drill. Endpoint under study: CLI faq_ask.

Capacity note for planners: estimate peak demand on CLI faq_ask and the cost ceiling for a failed retry storm. Write the abort conditions — unbounded spend, cross-tenant leakage, or inability to roll back — before you enjoy the first green test. Prefer synthetic fixtures shaped like production over anonymized production dumps you cannot share in class. When you are tempted to widen scope, re-read the oracle (gold 5/5 cited + 3/3 abstain; demo under 120s from clean checkout) and cut features that do not serve it. The teaching outcome is judgment under constraints: café ops lead answering hours/wifi/pets questions from a kiosk gets a trustworthy control, not a kitchen-sink framework. Keep the language of release decisions: promote, hold, or roll back — never “see if it gets better.”

Decide what will live in version control on day one: fixtures, contract markdown, and a failing test name. Write the cost ceiling as a hard number with currency and period. If the lab involves clusters, name the non-prod context you will use and forbid prod kubecontexts in scripts. Capture the baseline metric once before changing code so later gains are comparative. Refuse tools that hide the request path behind magic macros until the oracle is green on fakes. In this chapter the product is grounded café FAQ app with validation, five-note retrieval, citations, tests, demo, the human stakeholder is café ops lead answering hours/wifi/pets questions from a kiosk, and the incident id you design against is CAP-CAFE-DEMO-01. Re-state the oracle in your notes — gold 5/5 cited + 3/3 abstain; demo under 120s from clean checkout — and keep the invariant visible: unsupported questions abstain; every answer cites a note id; two-minute clean-room demo. Track gold_pass and demo_seconds as the scoreboard. Surface under change control: CLI faq_ask.