CreateCollection
Build a new database collection in Astra DB MCP Server, specifying collection name and optional vector properties for efficient data organization and retrieval.
Instructions
Create a new collection in the database
Input Schema
TableJSON Schema
| Name | Required | Description | Default |
|---|---|---|---|
| collectionName | Yes | Name of the collection to create | |
| dimension | No | The dimensions of the vector collection, if vector is true | |
| vector | No | Whether to create a vector collection |
Implementation Reference
- tools/CreateCollection.ts:19-47 (handler)The core handler function that implements the CreateCollection tool logic. It creates a collection in Astra DB, optionally configuring it as a vector collection with specified dimensions and metric.
export async function CreateCollection(params: { collectionName: string; vector?: boolean; dimension?: number; metric?: VectorMetric; }) { const { collectionName, vector = true, dimension = 1536, metric = "cosine" as VectorMetric } = params; if (vector) { await db.createCollection(collectionName, { vector: { dimension: dimension, metric: metric }, }); } else { await db.createCollection(collectionName); } return { success: true, message: `Collection '${collectionName}' created successfully`, }; } - tools.ts:54-84 (schema)The JSON schema definition for the CreateCollection tool input, including parameters like collectionName (required), vector, dimension, and metric with defaults.
{ name: "CreateCollection", description: "Create a new collection in the database", inputSchema: { type: "object", properties: { collectionName: { type: "string", description: "Name of the collection to create", }, vector: { type: "boolean", description: "Whether to create a vector collection", default: true, }, dimension: { type: "number", description: "The dimensions of the vector collection, if vector is true", default: 1536, }, metric: { type: "string", description: "The similarity metric to use for vector search (cosine, euclidean, or dot_product)", default: "cosine", enum: ["cosine", "euclidean", "dot_product"] } }, required: ["collectionName"], }, }, - index.ts:84-98 (registration)The registration and dispatch logic in the MCP server's CallToolRequest handler. It extracts arguments and calls the CreateCollection handler function, returning the result as MCP content.
case "CreateCollection": const createResult = await CreateCollection({ collectionName: args.collectionName as string, vector: args.vector as boolean | undefined, dimension: args.dimension as number | undefined, metric: args.metric as "cosine" | "euclidean" | "dot_product" | undefined, }); return { content: [ { type: "text", text: createResult.message, }, ], };