Ship and explain the café FAQ retriever

A shippable lab artifact includes: how to run it, the metric result (retrieval hit rate, citation correctness, supported-answer rate, abstention accuracy), the failure you can still reproduce (retrieval miss or ungrounded fluent answer), the security gate for prompt injection inside a stored FAQ note treated as instructions, and a rollback note. The user decision it supports remains: answer a café question only when a relevant local note is retrieved and cite its ID.

release={'docs_version':'faq-2026-07-18','gold_pass':3,'gold_total':3,'owner':'cafe-ops','rollback':'faq-2026-07-01'}
assert release['gold_pass']==release['gold_total'] and release['rollback']
print(release)

Expected evidence: stage evidence for mastery ship. Store this beside the fixture version so scores remain meaningful after content changes in build-mini-rag.

State operating limits for the café FAQ retriever in plain language: fixture size, offline vs live dependencies, and what would require a new eval set. Shipping build-mini-rag is honest scoping, not maximal confidence language.

Fill this table in your notes for the café FAQ retriever:

• Proved on café notes: hours, wifi, pets, allergens, parking: …
• Unproved beyond the fixture: …
• Metric that blocks release: retrieval hit rate, citation correctness, supported-answer rate, abstention accuracy
• Failure still reproducible: retrieval miss or ungrounded fluent answer
• Security gate: prompt injection inside a stored FAQ note treated as instructions
• Rollback: …

Ship the narrative only when the unproved list is honest. Reviewers trust narrow claims that support answer a café question only when a relevant local note is retrieved and cite its ID more than maximal language that collapses under the first production oddity.

Hand your ship note to a peer and ask them to recreate a proved/unproved ship note with rollback without watching you type. If they cannot, your evidence is still tribal knowledge. Tighten the run command and the metric line until a stranger can validate the café FAQ retriever against café notes: hours, wifi, pets, allergens, parking.

At the mastery and shipping stage for build-mini-rag, the job is narrower than finishing a product demo. You are creating one progressive evidence piece about café notes: hours, wifi, pets, allergens, parking that later pages inherit without redefining success. Keep that fixture small enough to inspect by hand, keep outputs copy-pasteable as text, and refuse to narrate this baseline as if it were a production SLA: always-answer-from-largest-document heuristic.

For this page specifically, success looks like a proved/unproved ship note with rollback while still centering the user decision to answer a café question only when a relevant local note is retrieved and cite its ID. If you cannot point to a file, command, or assertion that proves that for the café FAQ retriever, stay on this page instead of advancing.

How-to: build a 5-document RAG app · Glossary: RAG

Previous