create_table

Create a new local table in a SAP Datasphere space.

IMPORTANT: This is a HIGH-RISK, WRITE operation. It creates a design-time object and (optionally) deploys it, which changes the state of the tenant. Requires user consent.

Use this tool when:

• User asks: "Lege eine Tabelle X in Space Y an" / "Create a table for orders in DIMITRITEST"

• Provisioning new local tables for data ingestion, staging, or modelling

• Bootstrapping schemas from a specification (name + columns)

What it does:

1. Builds a Datasphere object definition (CSN-style JSON) from your column list

2. Writes it to a temporary file

3. Calls the Datasphere CLI: datasphere objects local-tables create --space <id> --file-path <def.json> (the table name comes from the definitions key in the JSON, not from a flag)

4. The CLI deploys the object by default so it becomes queryable; deploy=false adds --no-deploy

5. Returns the created table metadata

Required parameters:

• space_id: Target space (must exist, uppercase, e.g., 'DIMITRITEST')

• table_name: Technical name (uppercase, alphanumeric + underscore, e.g., 'BESTELLUNGEN')

• columns: Array of column definitions

Column definition format:

{
  "name": "BESTELL_ID",          // required, uppercase
  "type": "NVARCHAR",            // required: NVARCHAR|VARCHAR|INTEGER|BIGINT|DECIMAL|DOUBLE|DATE|TIMESTAMP|BOOLEAN|NCLOB
  "length": 20,                  // for NVARCHAR/VARCHAR
  "precision": 15,               // for DECIMAL
  "scale": 2,                    // for DECIMAL
  "nullable": false,             // default true
  "description": "Order ID"      // optional
}

Optional parameters:

• primary_keys: Array of column names that form the primary key (e.g., ["BESTELL_ID"])

• label: Business-friendly display name (e.g., "Bestellungen")

• description: Description of the table's business purpose

• deploy: If true (default), deploy the table immediately after creation so it can be queried

Example call (Bestellungen):

{
  "space_id": "DIMITRITEST",
  "table_name": "BESTELLUNGEN",
  "label": "Bestellungen",
  "description": "Kundenbestellungen mit Positionen",
  "primary_keys": ["BESTELL_ID"],
  "columns": [
    {"name": "BESTELL_ID",   "type": "NVARCHAR", "length": 20, "nullable": false},
    {"name": "KUNDEN_ID",    "type": "NVARCHAR", "length": 20, "nullable": false},
    {"name": "BESTELLDATUM", "type": "DATE"},
    {"name": "PRODUKT_ID",   "type": "NVARCHAR", "length": 20},
    {"name": "MENGE",        "type": "INTEGER"},
    {"name": "EINZELPREIS",  "type": "DECIMAL", "precision": 15, "scale": 2},
    {"name": "GESAMTBETRAG", "type": "DECIMAL", "precision": 15, "scale": 2},
    {"name": "WAEHRUNG",     "type": "NVARCHAR", "length": 3},
    {"name": "STATUS",       "type": "NVARCHAR", "length": 20}
  ],
  "deploy": true
}

Prerequisites (real mode):

• Datasphere CLI installed and available on PATH (datasphere --version)

• User logged in to the CLI (datasphere login)

• User has DW Space Administrator or Modeler role in the target space

Mock mode (USE_MOCK_DATA=true):

• No CLI call is made

• Returns a simulated success response with the generated object definition

• Perfect for testing prompts before running against a real tenant

Security & Safety:

• Table name is validated to prevent injection (must match ^[A-Z][A-Z0-9_]*$)

• Column names are validated identically

• Data types are checked against a whitelist

• HIGH-RISK: requires consent, all actions are audit-logged

• Idempotency: fails if the table already exists (use datasphere objects local-tables update / delete first)

Note: Corresponds to CLI: datasphere objects local-tables create --space <id> --file-path <def.json> [--no-deploy]

Only local tables. This is the one object type the server creates, because the payload is derived from data it has already read. Views, analytic models, flows, task chains, spaces, users and roles are CLI territory — see docs/MCP_VS_CLI.md. Do not try to reshape this tool's CSN into another type.

Canonical CSN shape: the generated definition follows examples/local-table-orders.json and Appendix E of the datasphere-cli skill (DataphereCLI repo). That example is not just a reference — tests/test_csn_shape.py derives this tool's arguments from it and compares the result back, so the two cannot drift apart. If the CLI ever rejects the payload, compare against the example rather than guessing.

Column names must be UPPERCASE (^[A-Z][A-Z0-9_]*$). The minimal SAP Help example uses a mixed-case Name, which this tool rejects on purpose: uppercase matches the Open SQL schema convention and needs no quoting in later queries.