Vector search

Managed retrieval with a folded static 512-dimensional model, portable int8 indexes, and one API across exact Mojo CPU search and capability-verified cuVS CAGRA workers.

Create a retrieval project

The same mojojojo account and API key owns your source documents, indexes, usage, and credits. No separate database account.

Create account — $1 free Sign in

API

curl https://mojojojo.cc/v1/vector/projects/vec_PROJECT/upsert \
  -H "Authorization: Bearer mj_live_…" -H "Content-Type: application/json" \
  -d '{"documents":[{"id":"guide","text":"Mojo compiles fast kernels."}]}'

curl https://mojojojo.cc/v1/vector/projects/vec_PROJECT/search \
  -H "Authorization: Bearer mj_live_…" -H "Content-Type: application/json" \
  -d '{"query":"fast compiled code","limit":10}'

What the latency number means

The response reports the backend and the measured end-to-end worker latency for that query. Exact CPU is available now. Performance projects may move to cuVS CAGRA GPU search only after that worker passes its health and benchmark probe; the API never labels a CPU fallback as GPU.