Scaling
How Doram grows with your data and your team.
The retriever, compiler and analyst hold no state, so they scale out as more people ask questions.
Every read of the graph is bounded. It walks at most three steps and visits at most 2,000 nodes, so the cost of a read depends on the question, not on the size of the graph.
A workspace is designed for one million nodes and five million edges. Above that, it gets its own database and read replicas.
After the first run, background jobs only process what changed.