vibedata-dbt-fabric-notebook 0.1.0

VibeData dbt runtime — Fabric notebook execution unit: clone the domain repo (Key Vault m2m PAT) and run a list of dbt commands on the official dbt-fabricspark adapter, persisting run_results/manifest to the lakehouse. notebookutils is provided by the Fabric runtime; dbt + dbt-fabricspark are pip-installed by the notebook.

vibedata-dbt-fabric-notebook

Fabric notebook execution unit for dbt — the dbt counterpart of vibedata-dlt-fabric-notebook. Clones the domain repo and runs a list of dbt commands on the official dbt-fabricspark adapter (Livy/Spark), persisting run_results.json + manifest.json + dbt.log to the lakehouse (the dbt audit the runtime-audit read layer reads — the dbt analog of dlt's finalize()).

Use inside a Fabric notebook

%pip install vibedata-dbt-fabric-notebook dbt-fabricspark>=1.10 -q

from vibedata.dbt.fabric.notebook import clone_domain_repo, run_dbt_commands

repo = clone_domain_repo(REPO_URL, vault_url=VAULT_URL, branch=BRANCH)   # KV m2m PAT
run_dbt_commands(
    ["dbt deps", "dbt build --select +fct_orders --target prod"],
    project_dir=f"{repo}/transformation",
)

• Auth: the domain repo is cloned with the m2m GitHub secret (domain-github-pat-token) read from Key Vault via notebookutils — the same slot the dlt runner uses. No OneLake deploy key.
• Command list: dbt Core has no "run a list" concept; run_dbt_commands is a thin fail-fast loop. Put dbt deps in the list if you need it — the runner never runs it implicitly.
• Zero pip deps of its own: notebookutils is provided by the Fabric runtime; dbt + dbt-fabricspark are installed by the notebook cell.

Follow-up

clone_domain_repo here duplicates the dlt runner's clone. They should fold into one shared, dlt-free *-notebook-core helper — deferred to avoid a refactor of the working dlt packages (fabric-core currently pulls dlt, so it isn't a clean shared home). Tracked as a follow-up.