Redis Feature Store Demo
\n\n A small fraud-scoring feature store. Each user is one Redis hash\n at __KEY_PREFIX__{id} with a batch-materialized\n batch half (daily aggregates,\n 24-hour key-level EXPIRE) and a streaming\n streaming half (real-time\n signals, __STREAM_TTL__s per-field HEXPIRE).\n Inference reads any subset with one HMGET; batch\n scoring pipelines HMGET across N users.\n
Store state
\nMaterialize batch features
\nCalls HSET + EXPIRE for each user\n through one Predis pipeline — the whole batch\n ships in one round trip.
\n Drop the TTL to e.g. 30 s and watch entities disappear on\n schedule — the same thing that happens if a daily refresher\n fails.\n
\n \n \nStreaming worker
\nPicks __USERS_PER_TICK__ users per tick, writes the\n streaming features, applies HEXPIRE\n __STREAM_TTL__s per field. Pause it and the\n streaming fields drop out via per-field TTL while the batch\n fields stay populated.
Inference read (HMGET)
\nPick a user and a feature subset. One HMGET\n round trip returns whatever the model needs.
Feature subset
\n\n Tick to include in the HMGET. Per-field TTL is\n shown next to each field in the result table.\n
Pick a user and click Read features.
\nBatch scoring
\nPipelined HMGET across N random users via\n Predis pipelines. One network round trip for the\n whole batch.
(no batch read yet)
\nInspect one user
\nHGETALL plus per-field HTTL and\n key-level TTL. Useful for spotting which\n streaming fields have already expired.
(pick a user and click Inspect)
\n