README

Tool for benchmarking vector search with randomized vectors.

Logic

The tool connects to the Redis/Dragonfly instance.

1. Checks if the database has enough data (at least 50% of requested -n). If not, it flushes the DB and generates random vectors.
2. Checks if the index idx exists. If not, it creates it.
3. Runs concurrent search queries and reports latency/QPS.

Arguments

Flag	Default	Description
-n	50000	Number of vectors to populate if the DB is empty.
-q	1000	Total number of queries to run during the benchmark.
-t	8	Query threads. Number of concurrent workers sending queries.
-d	100	Vector dimension. Size of the float32 vectors.
-k	10	Top K. Number of nearest neighbors to retrieve per query.
-p	6379	Port of the server.
-h	localhost	Host of the server.

Run with -h (help) to see these defaults in the tool itself.