Skip to content


Switch branches/tags

Latest commit


Git stats


Failed to load latest commit information.
Latest commit message
Commit time

dbt-helper (beta)

NOTE: dbt-helper currently only works with dbt version >= 0.17.1. This is still in extremely-limited beta release. We'd love your help, testing it though! Until we're out of beta, please don't build this into anything production-touching as it will surely get broken sooner rather than later

dbt-helper is a command line tool that helps with developing dbt projects and managing data warehouses.

This repository is not formally associated with dbt and is not maintained by fishtown-analytics (the maintainers of dbt). If you have problems with one of these (dbt-helper) tools, please file an issue on this repository and do not bother the dbt maintainers about it.


pip install dbt-helper

NOTE: dbt-helper may not work with dbt when dbt is installed via homebrew. We are currently investigating options for fixes, but if you experience issues that might be the culprit.


dbt-helper (currently) has seven sub-commands:

  • compare: Compare the relations in your warehouse with those that dbt is managing. This is useful for identifying "stale" relations that are no longer being updated by dbt (like if, for example, you converted the model from materialized to ephemeral).
    • Note: dbt-helper compare will compare all schemas that are impacted by models in the models/ directory. There is (currently) no way to specify a single schema to compare.
  • bootstrap: Create starter "schema.yml" files for your project. This function helpfully generates boilerplate dbt-model files for you so you don't have to go through the copy/paste when you're developing a new model.
    • Note: this command will not over-write existing schema.yml files. It will default to printing templates to the console, but you can create new files by using the --write-files flag.
  • show-upstream: Inspect the dbt graph and show the relations that are "upstream" from (i.e., the "parents" of) the selected relation. Print to the terminal.
  • show-downstream: The same as show-upstream but in the other direction -- show dependents 'downstream' from (i.e., the "children" of) the selected relation
  • find: Find the compiled .sql file for a model by providing the model name only. You can also find the source or run .sql files for a model by using the appropriate flag. Useful when working in large dbt projects and you want to find files quickly wihout having to navigate a file tree.
  • open: Open the compiled .sql file for a model by providing the model name only. Works the same as find, but directly opens the file in your text editor.
  • retry-failed: Rerun models that errored or were skipped on your previous dbt run.

As one might hope, you can view the command line options directly from the tool by using the help functionality:

dbt-helper --help

Usage Details:


$ dbt-helper compare

Comparing local models to the database catalog. Checking schemas:
- dev_downstream
- dev_example
Warning: The following relations do not match any models found in this project:
TABLE "dev_example"."b"
VIEW "dev_downstream"."d"


$ dbt-helper bootstrap --schemas dev_example

Bootstrapping the following schemas:
- dev_example
Design for relation: dev_example.my_first_dbt_model
version: 2
  name: my_first_dbt_model
- columns:
  - name: id
  description: 'TODO: Replace me'

Design for relation: dev_example.b
version: 2
  name: b
- columns:
  - name: city
  - name: count
  description: 'TODO: Replace me'


$ dbt-helper show-upstream d



see show-upstream


# Find the compiled version of model
$ dbt-helper find my_model

# Same as above
$ dbt-helper find my_model --compiled

# Find the run version of the model
$ dbt-helper find my_model --run

# Find the source version of the model
$ dbt-helper find my_model --source

Understanding the flags:

  • The --compiled flag will find the relevant .sql file in the target/compiled directory. This file contains the compiled SELECT query.
  • The --run flag will find the relevant .sql file in the target/run directory. The run version of the model contains the compiled query wrapped in the DDL (i.e. CREATE) statements to materialize it in the warehouse.
  • The --source flag will find the relevant jinja-flavored .sql file in the models/ directory.

Without a flag, dbt-helper will find the compiled model.


# Open the compiled version of model
$ dbt-helper open my_model

# Same as above
$ dbt-helper open my_model --compiled

# Open the run version of the model
$ dbt-helper open my_model --run

# Open the source version of the model
$ dbt-helper open my_model --source

# Print model to STDOUT rather than opening in a text editor
$ dbt-helper open my_model --print

These flags work exactly the same was find above, so check there for more detail on these.

If dbt-helper is opening your file in the wrong text editor, update your $EDITOR environment variable to be the command you normally use to open a file in your text editor, e.g. vim, emacs, atom, code, or subl. If you would prefer to not update your $EDITOR env var, you can instead create a $DBT_HELPER_EDITOR environment variable for your command.

dbt-helper will try to use one of the following commands to open the file (in order of precedence):

  • $DBT_HELPER_EDITOR. If this is not set, then it will use:
  • $EDITOR: A standard environment variable used across a number of applications (e.g. git prompts launch in the editor specified here). If this is not set, then use:
  • "Open": Generic command that will open a file in the default application for the associated file type.


$ dbt-helper retry-failed
dbt run --models my_failed_model my_skipped_model
Running with dbt=0.13.0
Found 8 models, 20 tests, 0 archives, 0 analyses, 113 macros, 0 operations, 3 seed files, 0 sources

17:40:11 | Concurrency: 1 threads (target='dev')
17:40:11 |
17:40:11 | 1 of 2 START view model dev_example.my_failed_model.................. [RUN]
17:40:12 | 1 of 2 OK created view model dev_example.my_failed_model............. [CREATE VIEW in 0.65s]
17:40:12 | 2 of 2 START table model dev_example.my_skipped_model................ [RUN]
17:40:29 | 2 of 2 OK created table model dev_example.my_skipped_model........... [SELECT in 16.86s]
17:41:31 |
17:41:31 | Finished running 1 view models, 1 table models in 80.96s.


Install locally for development:

pip install -e .


We follow a few conventions in this repository.

Git stuff

Please name your branch names with a prefix denoting the type of change you're making. For example:


Please also follow generally accepted git best practices by using descriptive commit messages and squashing your commits (where possible) before submitting your PR (or after making any required changes).

Python formatting

We use black to format python code in this project. Please use it! (And consider using it for all of your other python projects while you're at it.)

Running the tests

Run the tests using your local environment:

tox -e dev -- --nocapture

Run the tests locally using docker:

docker-compose run test tox -- --nocapture

If you want to specify a particular test to run, you can pass in a path to an integration test to either of the above commands, e.g.:

tox -e dev -- --nocapture test/integration/001_compare_test/

Giving Thanks

  • Thanks to Claire Carroll for the open and retry-failed functions
  • Thanks to Drew Banin for invaluable discussion and code-review on early dbt-helper features.
  • Thanks to John Lynch for early beta-testing and feedback.
  • Thanks to Leon Tchikindas for the blog post (and code) inspiring this command-line graph builder.


A helper tool for dbt development and data warehouse management.







No packages published

Contributors 4