Proposal: Investigate Dosu installation on Datafusion repositories

[Edmondo Porcu <ed...@gmail.com>]

Hello,

I have recently discovered an AI Github Assistant called https://dosu.dev/
which helps with auto-labeling, issue-triage and QA.

- It is used successfully in other projects under the Apache Foundation,
most notably Airflow and Superset.
- The examples on their homepage for labeling and question answering come
from Apache projects, see
https://docs.dosu.dev/pages/features-auto-labeling#usage
https://docs.dosu.dev/pages/features-issue-triage-qa#usage
- It can be run in preview mode where only users who have access to the
dosu console see it in action, and users on Github do not see anything

Given the success in other Apache repos, maybe we should give it a shot,
starting from only one repo?

Best
Edmondo

Re: Proposal: Investigate Dosu installation on Datafusion repositories

[Edmondo Porcu <ed...@gmail.com>]

There are three features at the moment:
-  Auto-labeling would be very helpful
https://docs.dosu.dev/pages/features-auto-labeling . Can't really test
because I don't have enough issues with label. This feature is used by
Superset and Airflow on their repo
- Auto-triage and QA would not be very helpful, since the average
contributor / participant to the Datafusion community is very technical,
and I don't think we have a lot of issues we could answer automatically
(unless someone is already closing the).
https://docs.dosu.dev/pages/features-issue-triage-qa
- Changelog generation could be very helpful
https://docs.dosu.dev/pages/features-changelog-generation . This feature is
used by Superset

I think it would be worth it to make this experiment and install Dosu to
the main repo in "Response Preview Mode", see
https://docs.dosu.dev/pages/response-previews. In this way only people
having access to dosu.dev workspace could see the answers, and observe the
tool maybe for 2 weeks on real user activity (real issues)

On Thu, Jan 23, 2025 at 2:11 PM Piotr Findeisen <pi...@gmail.com>
wrote:

> Hi
>
> What's your perceived quality of the "help"?
> I hope we agree it needs to be much much over "50% right" to be overall
> useful.
>
> PF
>
>
>
> On Tue, 21 Jan 2025 at 21:56, Edmondo Porcu <ed...@gmail.com>
> wrote:
>
> > I have enabled Dosu on my own fork of Datafusion, you can see how it
> > interacted with three issues I copied from the main repo:
> >
> > https://github.com/edmondop/arrow-datafusion/issues/6
> > https://github.com/edmondop/arrow-datafusion/issues/7
> > https://github.com/edmondop/arrow-datafusion/issues/8
> >
> > I also noticed that on enhancement issues, it hasn't replied, but it did
> on
> > bugs.
> >
> > Ed
> >
> > On Sun, Jan 5, 2025 at 12:58 PM Andrew Lamb <al...@influxdata.com>
> wrote:
> >
> > > Sounds like an interesting idea to try in my opinion. Thanks for
> bringing
> > > it up Edmondo
> > >
> > > On Sun, Jan 5, 2025 at 11:48 AM Edmondo Porcu <edmondo.porcu@gmail.com
> >
> > > wrote:
> > >
> > > > Hello,
> > > >
> > > > I have recently discovered an AI Github Assistant called
> > > https://dosu.dev/
> > > > which helps with auto-labeling, issue-triage and QA.
> > > >
> > > > - It is used successfully in other projects under the Apache
> > Foundation,
> > > > most notably Airflow and Superset.
> > > > - The examples on their homepage for labeling and question answering
> > come
> > > > from Apache projects, see
> > > > https://docs.dosu.dev/pages/features-auto-labeling#usage
> > > > https://docs.dosu.dev/pages/features-issue-triage-qa#usage
> > > > - It can be run in preview mode where only users who have access to
> the
> > > > dosu console see it in action, and users on Github do not see
> anything
> > > >
> > > > Given the success in other Apache repos, maybe we should give it a
> > shot,
> > > > starting from only one repo?
> > > >
> > > > Best
> > > > Edmondo
> > > >
> > >
> >
>

Re: Proposal: Investigate Dosu installation on Datafusion repositories

[Piotr Findeisen <pi...@gmail.com>]

Hi

What's your perceived quality of the "help"?
I hope we agree it needs to be much much over "50% right" to be overall
useful.

PF



On Tue, 21 Jan 2025 at 21:56, Edmondo Porcu <ed...@gmail.com> wrote:

> I have enabled Dosu on my own fork of Datafusion, you can see how it
> interacted with three issues I copied from the main repo:
>
> https://github.com/edmondop/arrow-datafusion/issues/6
> https://github.com/edmondop/arrow-datafusion/issues/7
> https://github.com/edmondop/arrow-datafusion/issues/8
>
> I also noticed that on enhancement issues, it hasn't replied, but it did on
> bugs.
>
> Ed
>
> On Sun, Jan 5, 2025 at 12:58 PM Andrew Lamb <al...@influxdata.com> wrote:
>
> > Sounds like an interesting idea to try in my opinion. Thanks for bringing
> > it up Edmondo
> >
> > On Sun, Jan 5, 2025 at 11:48 AM Edmondo Porcu <ed...@gmail.com>
> > wrote:
> >
> > > Hello,
> > >
> > > I have recently discovered an AI Github Assistant called
> > https://dosu.dev/
> > > which helps with auto-labeling, issue-triage and QA.
> > >
> > > - It is used successfully in other projects under the Apache
> Foundation,
> > > most notably Airflow and Superset.
> > > - The examples on their homepage for labeling and question answering
> come
> > > from Apache projects, see
> > > https://docs.dosu.dev/pages/features-auto-labeling#usage
> > > https://docs.dosu.dev/pages/features-issue-triage-qa#usage
> > > - It can be run in preview mode where only users who have access to the
> > > dosu console see it in action, and users on Github do not see anything
> > >
> > > Given the success in other Apache repos, maybe we should give it a
> shot,
> > > starting from only one repo?
> > >
> > > Best
> > > Edmondo
> > >
> >
>

Re: Proposal: Investigate Dosu installation on Datafusion repositories

[Edmondo Porcu <ed...@gmail.com>]

I have enabled Dosu on my own fork of Datafusion, you can see how it
interacted with three issues I copied from the main repo:

https://github.com/edmondop/arrow-datafusion/issues/6
https://github.com/edmondop/arrow-datafusion/issues/7
https://github.com/edmondop/arrow-datafusion/issues/8

I also noticed that on enhancement issues, it hasn't replied, but it did on
bugs.

Ed

On Sun, Jan 5, 2025 at 12:58 PM Andrew Lamb <al...@influxdata.com> wrote:

> Sounds like an interesting idea to try in my opinion. Thanks for bringing
> it up Edmondo
>
> On Sun, Jan 5, 2025 at 11:48 AM Edmondo Porcu <ed...@gmail.com>
> wrote:
>
> > Hello,
> >
> > I have recently discovered an AI Github Assistant called
> https://dosu.dev/
> > which helps with auto-labeling, issue-triage and QA.
> >
> > - It is used successfully in other projects under the Apache Foundation,
> > most notably Airflow and Superset.
> > - The examples on their homepage for labeling and question answering come
> > from Apache projects, see
> > https://docs.dosu.dev/pages/features-auto-labeling#usage
> > https://docs.dosu.dev/pages/features-issue-triage-qa#usage
> > - It can be run in preview mode where only users who have access to the
> > dosu console see it in action, and users on Github do not see anything
> >
> > Given the success in other Apache repos, maybe we should give it a shot,
> > starting from only one repo?
> >
> > Best
> > Edmondo
> >
>

Re: Proposal: Investigate Dosu installation on Datafusion repositories

[Andrew Lamb <al...@influxdata.com>]

Sounds like an interesting idea to try in my opinion. Thanks for bringing
it up Edmondo

On Sun, Jan 5, 2025 at 11:48 AM Edmondo Porcu <ed...@gmail.com>
wrote:

> Hello,
>
> I have recently discovered an AI Github Assistant called https://dosu.dev/
> which helps with auto-labeling, issue-triage and QA.
>
> - It is used successfully in other projects under the Apache Foundation,
> most notably Airflow and Superset.
> - The examples on their homepage for labeling and question answering come
> from Apache projects, see
> https://docs.dosu.dev/pages/features-auto-labeling#usage
> https://docs.dosu.dev/pages/features-issue-triage-qa#usage
> - It can be run in preview mode where only users who have access to the
> dosu console see it in action, and users on Github do not see anything
>
> Given the success in other Apache repos, maybe we should give it a shot,
> starting from only one repo?
>
> Best
> Edmondo
>