How it scales.

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 agent1,00010,000100,0001,000,000
Disk: memories / search index0.3 MB / 3.1 MB3 MB / 31 MB30 MB / 311 MB300 MB / 3.1 GB
RAM with the index loaded1.4 MB11 MB99 MB1.1 GB
Search, store kept open (Go)1.1 ms1.9 ms15 ms143 ms
Search per cheesedb command or MCP call9.0 ms79 ms0.74 s7.5 s
Recall with Erase per command or MCP call31 ms112 ms0.81 s7.5 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.

Faster at recall and reads. Much slower to start.

The same data on all four engines (including Neo4j, a graph database), with every write flushed to the physical disk before it is acknowledged.

Recall that removes 10,000 memories

CheeseDB5.2 ms
SQLite236 ms
PostgreSQL1.94 s
Neo4j68 ms

Log scale.

Read one memory by id

CheeseDB1.8 µs
SQLite7.4 µs
PostgreSQL59 µs
Neo4j211 µs

Log scale.

1M memories per agentCheeseDBSQLitePostgreSQLNeo4j
Recall that removes 10,000 memories5.2 ms236 ms1.94 s68 ms
Recall that removes 20 memories4.9 ms7.2 ms10.1 ms6.5 ms
Read one memory by id1.8 µs7.4 µs59 µs211 µs
Durable write (one memory)4.6 ms5.5 ms6.8 ms6.1 ms
Startup, ready for the first query706 ms0.1 ms3.0 ms1.2 ms¹
Disk per memory314 B249 B546 B496 B
RAM per memory while open205 Bpage cache onlyshared buffers onlyJVM 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.

Where these numbers come from.

Every number on this page can be checked, and repeated.