One agent with a search index (768 values per memory), from a thousand memories to a million. These are CheeseDB’s own times: the embedding model is not included. With a local one, turning a question into a vector took 17 ms.
| Memories in one agent | 1,000 | 10,000 | 100,000 | 1,000,000 |
|---|---|---|---|---|
| Disk: memories / search index | 0.3 MB / 3.1 MB | 3 MB / 31 MB | 30 MB / 311 MB | 300 MB / 3.1 GB |
| RAM with the index loaded | 1.4 MB | 11 MB | 99 MB | 1.1 GB |
| Search, store kept open (Go) | 1.1 ms | 1.9 ms | 15 ms | 143 ms |
Search per cheesedb command or MCP call | 9.0 ms | 79 ms | 0.74 s | 7.5 s |
| Recall with Erase per command or MCP call | 31 ms | 112 ms | 0.81 s | 7.5 s |
Median of 3 runs on an Apple M1 Max. In memory each vector takes one byte per value; the few rows that could make the top results are re-checked with the exact values on disk, so results are exact. The command and the MCP server open the store for every call, so each search reloads the index; they suit agents of up to about 100,000 memories. A Go program that keeps the store open loads it once. With real embeddings and varied text at 1M, search still ranked the right memories first among a million look-alikes, and none came back after their source was recalled. Models, measured apart: embedding one question 17 ms; embedding a million memories about 36 minutes (CheeseDB’s part: 42 s). Method and every number: BENCHMARKS.md.
The same data on all four engines (including Neo4j, a graph database), with every write flushed to the physical disk before it is acknowledged.
| 1M memories per agent | CheeseDB | SQLite | PostgreSQL | Neo4j |
|---|---|---|---|---|
| Recall that removes 10,000 memories | 5.2 ms | 236 ms | 1.94 s | 68 ms |
| Recall that removes 20 memories | 4.9 ms | 7.2 ms | 10.1 ms | 6.5 ms |
| Read one memory by id | 1.8 µs | 7.4 µs | 59 µs | 211 µs |
| Durable write (one memory) | 4.6 ms | 5.5 ms | 6.8 ms | 6.1 ms |
| Startup, ready for the first query | 706 ms | 0.1 ms | 3.0 ms | 1.2 ms¹ |
| Disk per memory | 314 B | 249 B | 546 B | 496 B |
| RAM per memory while open | 205 B | page cache only | shared buffers only | JVM heap and page cache |
Median of 3 runs on an Apple M1 Max (macOS, internal SSD), SQLite 3.53, PostgreSQL 17 and Neo4j Community 2026.10. PostgreSQL and Neo4j ran over TCP on localhost. Writes are close because the disk flush dominates for every engine. SQLite is smaller on disk and starts far faster. ¹ A new connection to a server that is already running; booting the Neo4j server took 6.5 to 8.2 s in an earlier run. The Neo4j model and queries are ours, and an expert may do better. These numbers are for the library without a search index. Method and caveats are in BENCHMARKS.md.
Every number on this page can be checked, and repeated.
bench/run-all.sh (how to reproduce).