Vector Index

This chapter introduces the indexes supported in Hippo. Please kindly check the table listed below. Hippo supports building multiple vector indexes on the same vector field.

IndexesTypesApplication Scenarios
FLAT / BIN_FLAT
  • Small datasets
  • Requires 100% recall rate
  • IVF_FLAT / BIN_IVFQuantization-based Index
  • High-performance query
  • Recall rate as high as possible
  • IVF_SQQuantization-based Index
  • High-performance query
  • Limited memory resources
  • Higher recall rate
  • IVF_PQQuantization-based Index
  • Ultra high-performance query
  • Limited memory resources
  • Lower recall rate than IVF_SQ
  • IVF_PQ_FSQuantization-based Index
  • Ultra high-performance query
  • Limited memory resources
  • Lower recall rate than IVF_SQ
  • HNSW / BIN_HNSWGraph-based Index
  • Ultra high-performance query
  • Large memory resources
  • Recall rate as high as possible
  • Table 8 Vector Index Supported in Hippo