Typeahead / autocomplete

Return the top k query completions on every keystroke in under 50 ms so suggestions feel instantaneous. The trick is precomputing the top-k list at every trie node offline or in a streaming window, serving reads entirely from memory, and never mutating the trie synchronously on the read path.

Notes and ratings save in this browser. Completing self-assessment records your progress.

Duration 45:00

Requirements checklist

As the user types a prefix, return the top k (typically 5–10) most relevant completions ranked by popularity and recency.Record completed searches and fold trending queries into suggestions within minutes.Filter out offensive, sensitive, or legally blocked phrases before serving.Server-side p99 latency under 50 ms (total round-trip under 100 ms so it stays ahead of typing speed).Read-dominated workload (~10× more prefix lookups than completed searches); high availability via in-memory replicas.Eventual consistency for frequency updates is fine, except blocklist removals which must take effect quickly.

Clarifying questions

How many suggestions per prefix, and are they global, per-language, or personalised?
How quickly must a newly trending query appear in suggestions?
Do we support spell correction / fuzzy prefix matching or strict prefixes only?
How fast must a blocked/offensive suggestion be removed globally?

Open a practice canvas (new tab)Estimation calculators (new tab)

requirements

Suggested time: about 6 minutes. Separates keystroke read path from completed-query logging and trending velocity.

Notes for requirements