Home▸AI & Machine Learning▸Search Index Freshness & TTL Simulator

Search Index Freshness & TTL Simulator

Watch a retrieval index in 3D as documents arrive, sit in an ingestion queue, go live, and expire on a TTL clock. Tune arrival rate, ingestion latency, TTL window and topic count while tracking live freshness hit rate, index coverage and staleness.

AI & Machine Learning3DModerate60 FPS📱 Mobile-adapted⇄ 2D version
Search Index Freshness & TTL Simulator ↗ Open standalone

Every retrieval-augmented system faces the same tension once documents start changing: an index that never updates goes stale, but re-indexing everything constantly is expensive. This simulator models a live retrieval index as a ring buffer of document slots on a rotating torus — new documents spawn as a Poisson arrival process, sit in an outer "staging" ring while an incremental indexer catches up to an ingestion-latency delay, then join the inner "indexed" ring where they age from fresh green toward TTL-orange until either the TTL clock or capacity pressure evicts them. A simulated query stream continuously probes a set of topics and measures whether each topic's most current document is actually live in the index right now, giving three real-time metrics — freshness hit rate, index coverage and average indexed-document age — that trade off directly against the arrival rate, ingestion latency, TTL window and topic-count controls.

⚙ Under the hood

Watch a retrieval index in 3D as documents arrive, queue for ingestion, go live, and expire on a TTL clock. Tune arrival rate, ingestion latency, TTL window and topic count while tracking live freshness hit rate, index coverage and staleness.

AIInformation RetrievalRAGSearch IndexTTLThree.js

3D · Three.js / WebGL renderer · 60 FPS target · runs fully client-side, no install

What did you find?

Add reproduction steps (optional)