Taruvi app developer (backend)

Single backend skill for Taruvi. Covers two layers:

• Control plane (MCP) — provisioning datatables, policies, roles, buckets, secrets, function metadata, analytics queries, audited raw SQL.
• Data plane (function runtime) — Python bodies with signature def main(params, user_data, sdk_client) running on Celery workers with a pre-authenticated sdk_client.

If the task is wiring React/Refine providers or hooks on the frontend, switch to taruvi-refine-providers.

MCP tool names

This skill describes MCP operations ("save a table schema", "run a function"), not tool names — the server's tool names and argument shapes change over time. Pick the tool from the connected server's tool list whose description matches the operation, and follow that tool's description and input schema. When the live tool description disagrees with this skill, the tool description wins. If no tool covers an operation, say so and use the SDK, REST API, or Console instead of guessing a name.

Decision: function or just MCP provisioning?

Provisioning alone is enough for: schema changes, policy/role/user management, bucket creation, secret management, registering analytics queries, one-shot admin SQL.

You need a function body when the task involves any of: