Managed Indexes
- HNSW, IVF-PQ and DiskANN
- Shard and replica planning
- Rebuilds with no read downtime
- Per-query latency headers
You give us a recall target and a latency budget. We choose the index family, the shard count and the replica layout, and we keep them there as the collection grows.
Somebody else's full-time job
Nodeform picks HNSW under fifty million vectors, DiskANN above it, and moves you between them without a migration weekend. Rebuilds run on a shadow replica and cut over atomically, so reads never stop.
Why a managed index is worth it
A number, not a promise
24 ms p99 on the Scale tier, measured on your own traffic and reported weekly. Two weeks above it credits the month back.
No migration weekends
Index family changes, shard splits and dimension changes all happen behind a shadow replica. Your reads see one atomic cutover.
Capacity before the cliff
Memory headroom is watched per shard. You get the resize before p99 moves, not the incident after it.
Ready to stop running
your own vector index?
Read the quickstart, or talk to the engineer who will run your index.
