Skip to content

FAQ

What is DQL?

An open-source analytics language, CLI, and notebook. Every analytics answer — SQL, chart config, owner, tests, parameters — lives in a single .dql file in git. See Concepts.

Do I need a cloud warehouse to try it?

No. DQL ships with DuckDB. Drop a CSV into data/, open dql notebook, and query it immediately. No credentials, no configuration.

Do I need dbt?

No. DQL works standalone. But if you already use dbt, create-dql-app auto-detects a sibling dbt project. Run dbt build, then dql compile ., dql sync dbt ., and dql agent reindex to refresh manifest metadata, semantic metrics, lineage, and the agent index. See Import a dbt project.

How is this different from a Jupyter notebook?

Jupyter DQL notebook
Primary language Python SQL
Runtime Python kernel DuckDB (or your warehouse)
Chart setup matplotlib/Plotly code declarative, auto-rendered
File format .ipynb JSON .dqlnb JSON
Git diff noisy (outputs embedded) clean (outputs sidecar)

How is this different from Metabase, Looker, or Hex?

Those are BI tools — queries live in a database or cloud service. DQL keeps every artifact in git. You get code review, diff, branching, and blame on your analytics. See Migrate from Metabase / Looker / Hex.

What's the difference between custom and semantic blocks?

  • type = "custom" — the block's SQL executes directly against a connection
  • type = "semantic" — the block references a semantic-layer metric; SQL is generated

Start with custom blocks if you're new.

Can I use DQL from code, not just the CLI?

Yes. The useful entry points:

  • @duckcodeailabs/dql-core — parser, AST, formatter, semantic analysis
  • @duckcodeailabs/dql-compiler — IR lowering, codegen
  • @duckcodeailabs/dql-runtime — browser runtime
  • @duckcodeailabs/dql-connectors — database drivers

Does certification hit a real database?

Yes. dql certify <file.dql> --connection <driver> runs block assertions against the configured connection. In generated OSS projects, prefer dql validate for structural checks and dql certify for the local trust gate.

What's not in the open-source repo?

Hosted multi-user workspaces, governed secrets, audit logs, organization memory, approval workflows, managed alerting, regulatory governance packs, and permissions-aware team retrieval. See Compatibility.

Where do I report bugs or request features?

github.com/duckcode-ai/dql/issues.