Benchmark
10,000 memories · measured
Seed 10,000 mock memories with real 384-dim embeddings, then run the hybrid search pipeline against them to measure end-to-end latency. Every number you see below was recorded live on this machine, against the production Supabase.
Rows in benchmark DB
0
target 10,000
Seed progress
0%
idle
Embedding model
loading
MiniLM-L6-v2
Queries in suite
10
mix of lexical + semantic
⚙️
Benchmark DB is empty
Click Seed 10K rows above to populate the benchmark table with 10,000 mock memories (real 384-dim embeddings, chunked inserts, ~15 seconds). Then click Run benchmark to measure p50/p95/p99 latency across 10 test queries.