How to add vector index when you import data from your azure sql database?

Hello Suraj !

Thank you for posting on Microsoft Learn Q&A.

You make your search index contain a vector field and you populate that field with embeddings either pre-computed in SQL or computed by a skillset during indexing. The vector index size 0 bytes in your screenshot means that your index has no vector field or it’s empty.

You can add a vector field with profile to your index :

{
  "name": "azuresql-index",
  "fields": [
    { "name": "id", "type": "Edm.String", "key": true, "filterable": true },
    { "name": "content", "type": "Edm.String", "searchable": true },
    { "name": "embedding", "type": "Collection(Edm.Single)", "searchable": true,
      "dimensions": 1536,                   
      "vectorSearchProfile": "v-hnsw" }
  ],
  "vectorSearch": {
    "algorithms": [{ "name": "hnsw", "kind": "hnsw" }],
    "profiles":   [{ "name": "v-hnsw", "algorithm": "hnsw" }]
  }
}

and create a skillset that generates embeddings from your text column and maps them to /document/embedding:

{
  "name": "azuresql-embed-skillset",
  "skills": [
    {
      "@odata.type": "#Microsoft.Skills.Text.EmbeddingSkill",
      "name": "make-embeddings",
      "context": "/document",
      "inputs": [{ "name": "text", "source": "/document/content" }],
      "outputs": [{ "name": "embeddings", "targetName": "embedding" }],
      "modelVersion": "text-embedding-3-large"   // or your model
    }
  ]
}

then wire your indexer to the skillset and map the output:

{
  "name": "azuresql-indexer",
  "dataSourceName": "azuresql-datasource",
  "targetIndexName": "azuresql-index",
  "skillsetName": "azuresql-embed-skillset",
  "outputFieldMappings": [
    { "sourceFieldName": "/document/embedding", "targetFieldName": "embedding" }
  ],
  "fieldMappings": [
    { "sourceFieldName": "Id", "targetFieldName": "id" },
    { "sourceFieldName": "Content", "targetFieldName": "content" }
  ]
}

Run the indexer you will see the vector index size will grow from 0 once embeddings are stored.