I created a first proof of concept of an integrati...
# plugins-general
j
I created a first proof of concept of an integration between Meltano and Dagster. It automatically generates all Software Defined Assets from the taps, and dbt models and shows this in the UI. This also allows you to easily run subsets of your data pipeline. In this example you can easily run all upstream assets that are used as input for the machine learning model (across different technologies). I think it would also be cool to connect your dbt models to your BI tooling, to get the whole dependency graph from top to bottom 🤔
a
How does it work? And can it run a subset of a tap via adjusting the
select
config?
j
Yes, it injects the assets you selected using the
select
functionality.
Oh sorry, missed the first part of your comment. It works by using the SelectService in Meltano which returns all the selected streams from a tap. These are exposed to Dagster as Assets with a certain key, these keys correspond to the sources in DBT (this allows for the edge between singer streams and dbt models) When you materialize a stream, it uses Meltano to run the tap and target (with the capability of selecting a subset of streams). For the DBT part I'm using `dagster_dbt`which points to the dbt instance that is installed in Meltano.
a
We've talked for a while about adding JSON output option for
meltano select list
. I've logged the issue and am pasting here in case its helpful: https://github.com/meltano/meltano/issues/6657 This could be an alternative to calling
SelectService
directly.
b
@jules_huisman I am on the same path to create such integration, Would you be able to share the integration ? I have started a basic integration that reads just the
meltano.yml
file and builds core SDAs, but I am currently stuck on materializing part, if you are making your solution available it would be great help
s
I would also be very interested in learning more about how you did this.
Right now I’m running meltano via dagster, but it’s an op that runs a shell command, so dagster doesn’t really know much about the individual pipelines.
j
It is going to change in the future, but you can check out the current version here: https://github.com/quantile-development/dagster-meltano/tree/development. I am a bit busy at the moment, so I will probably continue development somewhere next week.