Pair your devices with a code and playback position follows you: pause on this device, hit resume on the other. Position is saved to the site every minute and on pause.
Open this panel on your other device and enter the same code.
Starts this lesson and continues through 12 more to the end of the course.
"Ask a question, get a chart" is the demo that sells Quick to executives. Whether it works in production depends entirely on the semantic layer underneath — and that layer was redesigned with the Quick Suite launch. If you learned QuickSight Q topics before late 2025, most of what you know is now the legacy path. This lesson covers the new model, the old one you'll inherit, and the cheaper alternative (dashboard Q&A) that often makes topics unnecessary.
"A Topic in Quick Sight is the multi-dataset semantic layer that brings multiple enriched datasets together into a unified data model… whether you're building analysis visuals or asking natural language questions through Amazon Quick chat." — Topics, marked "Applies to: Enterprise Edition" (verified 2026-08-26)
The two structural facts (same page + topics-create, verified 2026-08-26):
That second point is the architectural change to internalize: enrichment is reusable. Enrich a dataset once; every topic (and chat session) that touches it benefits. Under the old model the metadata was trapped inside each topic object.
Authoring workflow (topics-create, verified 2026-08-26): Data → Topics tab → Create topic → add up to 12 datasets → define relationships → optional custom instructions → Publish. Prerequisites: "Amazon Quick Enterprise Edition enabled with Author or Admin role", datasets in SPICE or a supported Direct Query source.
Topics in chat (verified 2026-08-26) documents the pipeline in four steps: intent parsing → relationship traversal → SQL generation → result presentation, "with the generated SQL available for inspection in the Explanation panel."
And the SQL it can generate is real SQL: "Inner, left, right, and full outer joins; Union queries; Subqueries for negation patterns; Cross-grain comparisons; Self-joins for recursive hierarchies."
Compare the legacy engine (legacy-topics, verified 2026-08-26): an "ML-based fuzzy search model [that] selects one dataset, then queries only that dataset." One dataset per question vs runtime multi-dataset SQL — that's the upgrade in one sentence.
Same page (verified 2026-08-26):
The API tells the same story twice. The original operations —
CreateTopic, UpdateTopic, DescribeTopic, ListTopics, DeleteTopic, SearchTopics,
permissions ops, refresh-schedule ops, and reviewed answers
(BatchCreateTopicReviewedAnswer / BatchDeleteTopicReviewedAnswer / ListTopicReviewedAnswers) —
carry the legacy shape:
CreateTopic's
TopicDetails embeds per-dataset Columns (with ColumnSynonyms, CellValueSynonyms,
SemanticType), CalculatedFields, Filters (with FilterSynonyms), and NamedEntities
(verified 2026-08-26). The new model ships as a parallel V2 family:
aws quicksight create-topic-v2 | update-topic-v2 | describe-topic-v2 |
list-topics-v2 | delete-topic-v2 | search-topics-v2 |
describe-topic-permissions-v2 | update-topic-permissions-v2
CreateTopicV2's
body is just DataSets + DataSetRelations (left/right dataset ARN + column names) +
CustomInstructions (verified 2026-08-26) — the metadata is gone from the topic because it moved
to the datasets. When you automate topic provisioning (Q7 patterns), pick the family that matches
the topic generation you're managing; they are distinct object shapes, both live.
⚠️ Limit we could not verify: a fields-per-topic ceiling. It isn't on the topic pages or the Service Quotas table (checked 2026-08-26); the nearest documented number is "Data Prep: Fields per dataset — 2,000". If a customer asks, say that and point at the quota page rather than quoting folklore from the Q era.
Dashboard Q&A (verified 2026-08-26):
"Quick allows any Author to enable Q&A directly from their dashboards in one click without the need to create a Topic in Quick Sight… check the Allow data Q&A checkbox from the dashboard publishing menu."
Readers then get Ask a question about this dashboard. The page's comparison table says what a real topic still buys you: reviewed answers, custom Q&A metadata, autocomplete. And a cost note that finance will ask about: "dashboard Q&A is a feature that incurs the associated enablement fee."
Decision rule worth teaching: start with dashboard Q&A; graduate to a topic when you need curated answers, synonyms beyond what enrichment gives you, or cross-dataset questions.
Related toggles on the same publish modal family: Allow executive summary (executive summaries, verified 2026-08-26) — LLM-generated dashboard summaries, opt-in per dashboard.
create-topic. A colleague says "just add the new datasets."
What do you check first?create-topic manages the legacy shape (metadata inside
TopicDetails); multi-dataset topics are create-topic-v2 with DataSetRelations. The two
families are different object shapes.BatchCreateTopicReviewedAnswer or the console.