Storage and durability

Behind the API, Loom runs a storage engine that holds your memory and answers searches. You don’t manage any of it — but it helps to know what it gives you.

​

Your memory is safe

When you add something and the call returns, it’s saved for good. Loom writes every change to a log first and takes regular snapshots, so your memory survives restarts and failures.

​

It scales with your work

Loom keeps the memory you use most ready in fast storage and moves the rest to durable storage in the background. A workspace can grow well beyond what fits in memory without slowing down the things you touch often.

​

Search stays fast and correct

Loom keeps a compact index of your memory so searches stay quick as your data grows, with an exact fallback that guarantees you still get the right answer. Under heavy load it leans on the index more, and results stay correct — only the speed changes.

​

It understands connections

Your papers, claims, and sessions are stored as a network, not just a pile of text. So when a search finds a strong match, Loom can follow the links to bring in related pieces. The same connections power the clusters and importance scores you see in the graph view.

​

It remembers time

Every fact is stored with the period it was true for. That’s what lets you ask about an earlier point in time and lets Loom keep two conflicting facts side by side. See How memory works for how this shapes your results.