►
A
Good
afternoon,
everyone,
this
is
the
weekly
envelopes
demo
for
september
3rd
2021.
as
usual,
how
everything
that
we
do?
We
add
the
updates
of
weekly
on
this
issue
thread.
So,
if
you
want
to
know
what
happened,
what
is
happening
around
envelopes
at
gitlab
just
subscribe
to
this
issue
and
you'll
get
notifications.
A
A
The
two
highlights
over
here,
for
me,
were
their
biggest
pinpoint,
is
not
having
a
notebook
this
so
because
we
already
spoke
about
this
on
the
previous
updates,
but
jupiter
notebooks
are
very
weird
file
system.
They
are
json
file
with
lots
of
stuff
inside
it
makes
it
really
hard
to
have
diffs
and
code
reviews
and
things
like
that,
just
within
the
how
they
are,
but
it
is
possible
to
improve
them
and-
and
we'll
be
talking
about
this
a
little
bit
later.
A
But
this
is
one
of
the
major
pain
points
it's
for
this
customer.
It
was
a
a
larger
pinpoint
than
not
having
model
motor
tracking
or
pipelines.
No
notebook
diff
was
their
major
pain
point
and
I've
seen
this
a
lot
in
the
industry
around
already
this
customer
they
use
azure
databricks
databricks
is
a
very
large
envelopes
provider,
so
they
were
recently
evaluated
at
38
billion
dollars
upon
a
1.6
billion
valuation
at
serious
age.
So
it's
it's
a
very,
very
hot
market
right
now,
they're
also
with
gcp.
A
A
A
Allowing
them
to
have
full
reviews
will
help
them
get
closer
to
gitlab,
because
what
happens
for
data
scientists
that
they
just
use,
gitlab
or
github
or
whatever
tooling,
at
the
end
of
the
of
the
analysis
that
usually
take
two
three
weeks
and
they
make
one
commit,
and
it's
just
as
storage,
they
use
git
as
a
storage
for
work
done.
They
don't
use
git
gitlab
github
for
discussing
for
improving
for
quality.
A
Now,
it's
just
like
the
storage,
it's
where
they,
where
analysis
go
to
die.
Basically,
so
what
we
are
doing
here,
we
are
working
a
bit
right
now
on
some
small
quality
of
life
improvements,
so
to
get
warmed
up.
So,
for
example,
now
the
svg
images
graphs
they
have
the
the
labels
properly,
so
images
were
too
large
that
were
too
large
were
overflowing
and
you
couldn't
see
them,
and
now
they
are
correctly
being
reconciled.
A
But
beyond
that
we
are
also
working
on
the
diff.
I
think
it's.
This
is
extremely
important.
This
was
great
so
many
times
by
so
many
users.
I
myself
when
I
was
a
data
scientist.
This
was
one
of
my
major
pain
points
with
kit
and
gitlab,
and
our
strategy
here
right
now
is
we're.
Gonna
have
the
we
have
a
drupal
notebook
over
here?
A
What
are
we
gonna?
Do
we're
gonna,
convert
this
to
a
markdown
file
and
we're
gonna
dip
the
markdown
file,
so
this
conversion
is
the
user
doesn't
see
this,
but
when
we
show
the
merge
request,
the
diffs
are
going
to
be
on
the
markdown
file,
an
example
of
the
markdown
file.
This
was
a
conversion
used
done
with
pandoc.
A
Pandoc
is
a
universal
converter.
It
converts
from
everything
to
everything
and
has
this
really
really
good:
conversion
for
jupiter,
notebooks,
intro
markdown
and
although
it's
quite
a
large
tool
and
a
little
bit
too
heavy
to
add
as
a
dependency
on
italy
and
things
like
that,
we
should
use
this
as
our
benchmark
almost
so.
For
example,
this
is
that
same
notebook
converted
into
a
a
markdown,
so
you
have
here
the
table
and
the
table
is
over
here.
It
looks
really
well
for
they
think
this
is
great.
A
So
what
we
did
we
started
to
work
on
this
repository
over
here
I
find
bhmd.
We
are
implementing
we're
trying
to
implement
the
same
things
but
in
gold
so
that
it
attaches
better
and
it's
very
small
so
that
it
doesn't
add
a
lot
of
overhead
into
the
code
base.
So
what
we
were
able
to
do
so
far
was
we
are
already
able
to
display.
A
We
create
array
a
markdown
without
the
specs
necessary,
with
all
the
inputs
already
done,
some
things
we
need
to
fix
here
and
there
we
also
need
to
work
a
little
bit
to
start
work
now
on
the
on
the
output,
how
to
display
each
one
of
the
output
types
and
we
don't
expect
to
be
able
to
display
everything
in
the
beginning.
We
want
to
get
a
little
bit
better
than
we
are
well.
A
Anything
is
a
bit
better
than
the
current
state,
because
it's
impossible
to
do
any
kind
of
code
review,
so
we're
gonna
evolve
little
by
little.
Now
we're
almost
done
with
a
quick
setup
of
this
tool
to
just
push
it
and
then
we're
going
to
start
integrating
and
speaking
with
our
colleagues
and
how
to
integrate
this
into
into
the
intricate
level
itself.
A
Additionally,
we
are
working
a
little
bit
with
with
capacitating
our
not
capacity
but
helping
creating
resources
so
that
our
sales
colleagues
can
have
better
conversations
about
envelopes
with
with
customers.
A
lot
of
questions
come
in.
Our
customers
are
starting
to
think
about
develops,
but
we
don't
want
our
sales
representatives,
our
sales
colleagues
they're.
A
Not
they
don't
feel
ready
to
have
those
conversations
that,
because
it's
really
new,
it's
not
the
usual
scope
of
of
gitlab
and
we
need
to
be
able
to
provide
a
little
bit
more
of
of
resources
to
them.
So
here
we
have
a
little
bit
of
problems
that
users
face
in
the
google
ad
on
on
the
gitlab
on
the
mlaps
space
and
the
components
that
would
be
used
to
solve
them.
We
don't
have
these
components
yet
with
gitlab.
A
We
will
also
be
working
soon,
like
I
said,
both
as
your
data
breaks,
both
azure
and
gcp
connect
with
mlflow,
and
I
think
it's
ml
flow
is
a
tool
to
track
two
to
track
models,
so
model
versioning,
which
one
should
be
available,
which
version
should
be
the
one
in
production,
let's
track
metrics
for
models,
so
it
uses
for
reviews
for
those
use
cases
and
it's
very
common,
and
it
should
a
little
bit
work
on
how
we
can
integrate
better.
So
I
will
be
doing
this
soon
as
well.