Qdrant · Schema

VectorStorageType

Storage types for vectors

AIArtificial IntelligenceVector Databases
View JSON Schema on GitHub

JSON Schema

qdrant-vectorstoragetype-schema.json Raw ↑
{
  "$schema": "https://json-schema.org/draft/2020-12/schema",
  "$id": "#/components/schemas/VectorStorageType",
  "title": "VectorStorageType",
  "description": "Storage types for vectors",
  "oneOf": [
    {
      "description": "Storage in memory (RAM)\n\nWill be very fast at the cost of consuming a lot of memory.",
      "type": "string",
      "enum": [
        "Memory"
      ]
    },
    {
      "description": "Storage in mmap file, not appendable\n\nSearch performance is defined by disk speed and the fraction of vectors that fit in memory.",
      "type": "string",
      "enum": [
        "Mmap"
      ]
    },
    {
      "description": "Storage in chunked mmap files, appendable\n\nSearch performance is defined by disk speed and the fraction of vectors that fit in memory.",
      "type": "string",
      "enum": [
        "ChunkedMmap"
      ]
    },
    {
      "description": "Same as `ChunkedMmap`, but vectors are forced to be locked in RAM In this way we avoid cold requests to disk, but risk to run out of memory\n\nDesigned as a replacement for `Memory`, which doesn't depend on RocksDB",
      "type": "string",
      "enum": [
        "InRamChunkedMmap"
      ]
    },
    {
      "description": "Storage in a single mmap file, not appendable Pre-fetched into RAM on load",
      "type": "string",
      "enum": [
        "InRamMmap"
      ]
    }
  ]
}