LAB / TOOL 01
Size the index.
See the assumptions.
A planning calculator for fixed-size chunking. Change an input to see its effect; export the estimate when you are ready.
YOUR ESTIMATE
Raw vectors only. Index structures, source text, metadata, replicas and serving infrastructure add storage and cost.
Calculated locally. Your inputs are not uploaded.
METHOD & LIMITATIONS
A model you can inspect.
Chunk count
For each average document: one chunk if its token count fits; otherwise 1 + ceil((tokens − chunk size) / (chunk size − overlap)). Multiply by document count.
Embedding input
Document tokens plus the repeated overlap for every additional chunk. Multiply by the assumed price per million input tokens.
Raw vectors
Total chunks × dimensions × bytes per value. Displayed in decimal MB or GB. Index overhead and compression are not modeled.
Use it for planning
Real document lengths vary. This estimates identical average-length documents with fixed-size chunking, no added metadata tokens and no retries. It is not a benchmark, hosting quote or quality prediction.