dbt¶

Scaffold a dbt project pre-wired to a Microsoft Fabric Data Warehouse using the dbt-fabric adapter.

No dbt installation is required to run these commands — fabric-dw generates all project files itself. A requirements.txt inside the scaffolded project lists the required pip packages (dbt-core, dbt-fabric) so you can install them in a separate environment when you are ready to run dbt.

Targets: Data Warehouse

CLI¶

dbt init¶

Targets: Data Warehouse

Scaffold a new dbt project directory connected to a Fabric Data Warehouse. The command creates the folder, writes dbt_project.yml, profiles.yml, requirements.txt, .gitignore, standard dbt model directories, a sample model, and a README. If git is on your PATH and the target folder is not already a git repository, git init is run automatically.

Security

When --auth sp (Service Principal) is used, profiles.yml emits Jinja2 env_var() placeholders ({{ env_var('AZURE_TENANT_ID') }} etc.) instead of literal secrets. You must set AZURE_TENANT_ID, AZURE_CLIENT_ID, and AZURE_CLIENT_SECRET in your environment before running dbt.

Usage

fdw [-w WORKSPACE] dbt init [OPTIONS] [ITEM] FOLDER

Arguments

Argument	Description
`ITEM`	Name or ID of the Fabric Data Warehouse item (optional if set via `fdw config set warehouse`).
`FOLDER`	Path to the folder to create. Must not exist (unless `--force` is passed).

Options

Option	Default	Description
`--project-name TEXT`	derived from `ITEM` name	dbt project name (sanitised: lowercase, non-alphanumeric chars replaced with `_`).
`--profile-name TEXT`	same as `--project-name`	dbt profile name written into `profiles.yml` and `dbt_project.yml`.
`--schema TEXT`	`dbo`	Default target schema for dbt models.
`--target TEXT`	`dev`	dbt target name inside the profile.
`--threads INTEGER RANGE`	`4`	Number of dbt threads (1–64).
`--auth [auto\\|CLI\\|ServicePrincipal\\|interactive\\|sp]`	derived from active credential mode	Authentication method. `interactive` is an alias for `CLI`; `sp` is an alias for `ServicePrincipal`.
`--profiles-dir [project\\|home]`	`project`	Where to write `profiles.yml`. `project` writes it next to `dbt_project.yml`; `home` merges it into `~/.dbt/profiles.yml` (backs up existing file first).
`--with-sources`	off	Introspect the live warehouse and generate a `_sources.yml` file listing all schemas and tables.
`--force`	off	Overwrite an existing non-empty directory.

Examples

# Minimal — uses configured default workspace and warehouse
fdw dbt init SalesWH ./my_dbt_project

# Explicit workspace via -w
fdw -w MyWorkspace dbt init SalesWH ./my_dbt_project

# Service Principal auth; write profiles.yml to ~/.dbt/
fdw -w MyWorkspace dbt init SalesWH ./sales_dbt \
  --auth sp --profiles-dir home

# Scaffold with live source introspection (auto-generates _sources.yml)
fdw -w MyWorkspace dbt init SalesWH ./sales_dbt --with-sources

# Force-overwrite an existing folder
fdw -w MyWorkspace dbt init SalesWH ./sales_dbt --force

Scaffolded layout

<FOLDER>/
├── .gitignore
├── README.md
├── dbt_project.yml
├── profiles.yml          # only when --profiles-dir project (default)
├── requirements.txt      # pip install -r requirements.txt
├── models/
│   ├── staging/
│   │   └── _sources.yml  # placeholder, or real entries with --with-sources
│   └── my_first_model.sql
├── seeds/
├── snapshots/
├── tests/
├── macros/
└── analyses/

MCP tools¶

Generate dbt-fabric project file contents pre-wired to a Microsoft Fabric Data Warehouse. Unlike the CLI dbt init command, these tools return file contents as text rather than writing files to disk, making them suitable for AI-assisted workflows where the AI agent writes the files.

Security

When authentication="ServicePrincipal" is used, the returned profiles_yml contains Jinja2 env_var() placeholders ({{ env_var('AZURE_TENANT_ID') }} etc.) rather than literal secrets. Never hard-code secrets into source-controlled files.

generate_dbt_profile¶

Targets: Data Warehouse

Return the contents of all files needed to bootstrap a dbt project that connects to a Fabric Data Warehouse. No files are written; the caller is responsible for persisting the returned strings.

Parameters:

workspace (str) — workspace name or GUID.
item (str) — Data Warehouse name or GUID.
project_name (str, optional) — dbt project name. Defaults to the warehouse display name (sanitised to lowercase + underscores).
profile_name (str, optional) — dbt profile name. Defaults to the project name.
schema (str, default "dbo") — default target schema for dbt models.
target (str, default "dev") — dbt target name.
threads (int, default 4) — number of dbt threads (1–64).
authentication (str, optional) — authentication method ("auto", "CLI", "ServicePrincipal"). Defaults to the MCP server's active credential mode.
with_sources (bool, default False) — when True, introspect the live warehouse and include all schemas/tables in the returned sources_yml.

Returns: object with:

profiles_yml (str) — contents for profiles.yml (write next to dbt_project.yml or into ~/.dbt/profiles.yml).
dbt_project_yml (str) — contents for dbt_project.yml.
sources_yml (str) — contents for models/staging/_sources.yml (placeholder or real entries when with_sources=True).
requirements_txt (str) — pip requirements listing dbt-core and dbt-fabric.
gitignore (str) — contents for .gitignore, pre-configured for dbt projects.