Limits

What’s not in DB Engine

Honest scope boundaries. If you need any of these, surface it — some are roadmap, others are intentionally out.

Read-only. No INSERT, UPDATE, DELETE, MERGE, TRUNCATE, DROP, GRANT, REVOKE, ALTER. Even hand-edited workbench SQL passes through the validator that rejects non-SELECT statements. If you need write-back, that’s a different feature — surface the request.
Single-connection per query. No three-part cross-database names, no Snowflake-style external-table references to other connections.

Saved queries / parameterized templates. The workbench history captures past queries but there’s no “save this as a named query” surface. Re-asking the same question and trusting the planner is the workflow today.
Stored procedures / functions. The validator rejects CREATE FUNCTION. If you need a pre-defined transformation, do it in a view in your warehouse and let the planner select from that.
Long-running queries. 30s driver-level statement timeout. Workloads beyond that should run in your warehouse’s scheduler, not Pollen8.
Streaming results. Result set is materialized in full before return. For very wide result sets, page via SQL (LIMIT … OFFSET) not via streaming.

dbt / orchestration. Pollen8 doesn’t orchestrate transformation pipelines. Use dbt / Airflow / Dagster / Prefect; Pollen8 reads the result.
ML / forecasting widgets. Predictive analytics is a different problem space; not in DB Engine.
Replacing your BI tool. Pollen8’s own BI satellite is built for in-app analyst Q&A, not for replacing Looker / Tableau / Sigma for executive dashboards.

Hard limits per query:

Per-tenant rate limits configurable under License.