embedding.put · Umwelt docs

Reference/Commands/embedding.put

embedding.putcommandsince 0.1

Stores vectors made by your model, or replaces them, in one transaction.

Command lineTypeScript.NET

Request

umwelt call story.umw embedding.put '{
  "projectId": "ball",
  "spaceId": "demo",
  "items": [
    {
      "docType": "record",
      "docId": "maid-staff",
      "textHash": "04d718cba11bc4995d03fbe82412f2bd2d6156acb6dc27042c297ed94a4d01f2",
      "vector": [0.12, 0.84, 0.53]
    },
    {
      "docType": "record",
      "docId": "butler-staff",
      "textHash": "5502f404e90632e72e7395a6f6d15dc5837dee8e9dbbfea82020527e4fffea1c",
      "vector": [0.9, 0.1, 0.42]
    }
  ]
}'

Request

const result = await client.embeddingPut({
  projectId: 'ball',
  spaceId: 'demo',
  items: [
    {
      docType: 'record',
      docId: 'maid-staff',
      textHash: '04d718cba11bc4995d03fbe82412f2bd2d6156acb6dc27042c297ed94a4d01f2',
      vector: [0.12, 0.84, 0.53],
    },
    {
      docType: 'record',
      docId: 'butler-staff',
      textHash: '5502f404e90632e72e7395a6f6d15dc5837dee8e9dbbfea82020527e4fffea1c',
      vector: [0.9, 0.1, 0.42],
    },
  ],
});

Request

var result = engine.EmbeddingPut(new PutEmbeddingsInput
{
    ProjectId = "ball",
    SpaceId = "demo",
    Items = new()
    {
        new EmbeddingItem
        {
            DocType = "record",
            DocId = "maid-staff",
            TextHash = "04d718cba11bc4995d03fbe82412f2bd2d6156acb6dc27042c297ed94a4d01f2",
            Vector = new() { 0.12, 0.84, 0.53 },
        },
        new EmbeddingItem
        {
            DocType = "record",
            DocId = "butler-staff",
            TextHash = "5502f404e90632e72e7395a6f6d15dc5837dee8e9dbbfea82020527e4fffea1c",
            Vector = new() { 0.9, 0.1, 0.42 },
        },
    },
});

Response

{
  "stored": 2,
  "stale": 0
}

Payload

Field	Type	Description
`projectId`required	string	The project to write to.
`spaceId`required	string	The embedding space.
`items`required	EmbeddingItem[]	One per document: `docType`, `docId`, the `textHash` the vector was made from, and the `vector`.

Result

Field	Type	Description
`stored`always	integer	Vectors stored.
`stale`always	integer	Vectors whose `textHash` is not the document’s: kept, and listed as pending until replaced.

Rules

Vectors are normalized. They must have the space’s dimensions, finite values, and not be all zeros. Search then gives the same order and scores on every platform.

A vector grants nothing. It only orders documents a read would see anyway.

Errors

VALIDATION_FAILEDA vector has the wrong dimensions or invalid values, or the document is not searchable (an Awareness, a Holding without text).

NOT_FOUNDThe project, the space or a document does not exist.

HANDSHAKE_REQUIREDOver the protocol, hello was not sent first.

Every code is listed in Errors.

Payload

Result

Rules

Errors

See also