Collections API

Collections are the primary organizational unit in SemaDB, similar to tables in relational databases. Each collection contains points with a consistent schema.

Creating Collections

Example: Creating a Collection with Vamana Index

ivvp = IndexVectorVamanaParameters(
    vectorSize = 128,
    distanceMetric = "cosine",
    searchSize = 75,
    degreeBound = 64,
    alpha = 1.2
)

ischema = Dict("embedding" => IndexSchemaValue(
    type = "vectorVamana",
    vectorVamana = ivvp
))

req = CreateCollectionRequest(id = "vectors", indexSchema = ischema)
create_collection(client, "user1", req)

Example: Creating a Collection with Text Index

text_params = IndexTextParameters(analyser = "standard")
ischema = Dict("description" => IndexSchemaValue(
    type = "text",
    text = text_params
))

req = CreateCollectionRequest(id = "text_data", indexSchema = ischema)
create_collection(client, "user1", req)

Example: Listing Collections

collections = list_collections(client, "user1")
for col in collections.collections
    println("Collection: $(col.id)")
end

Example: Getting Collection Details

collection = get_collection(client, "user1", "vectors")
println("Collection ID: $(collection.id)")
println("Shard count: $(length(collection.shards))")

Collection Schema

Each collection has a schema that defines how data is indexed. The schema is a dictionary mapping field names to index types.

Supported Index Types

• vectorVamana: Graph-based approximate nearest neighbor search
• vectorFlat: Exact search (brute force)
• text: Text search with semantic capabilities
• string: String field indexing
• stringArray: Array of strings indexing
• integer: Integer field indexing
• float: Floating-point field indexing

Example Schema

schema = Dict(
    "embedding" => IndexSchemaValue(
        type = "vectorVamana",
        vectorVamana = IndexVectorVamanaParameters(
            vectorSize = 128,
            distanceMetric = "cosine"
        )
    ),
    "description" => IndexSchemaValue(
        type = "text",
        text = IndexTextParameters(analyser = "standard")
    ),
    "category" => IndexSchemaValue(
        type = "string",
        string = IndexStringParameters(caseSensitive = false)
    )
)

Error Handling

When working with collections, you may encounter errors. The API returns appropriate error responses that you can handle:

try
    create_collection(client, "user1", CreateCollectionRequest(id = "invalid name", indexSchema = Dict()))
catch e
    println("Error creating collection: $(e)")
end

Common errors include:

• Invalid collection name format
• Schema validation errors
• Collection already exists
• Permission denied