►
A
A
Sure,
basically,
what
we
want
to
do
is
to
add
the
the
remote
development
module
to
to
the
gitlab
agent,
which,
as
of
right
now,
at
least
for
the
first
iteration,
we're
thinking
of
just
having
Agent
K
call
the
kitlab
rails
API
directly
through
a
helper
interface,
I
think
through
some
helper
struct
interface
functions
that
are
already
in
in
the
GitHub
agent.
So,
basically,
you
would
avoid
having
to
code
anything
for
the
server
side
for
the
cast
side
and
only
have
to
go
for
the
agent
K
side.
A
But
we
won't
get
away
with
is
having
no
optimization
whatsoever,
as
in
we
will
have
a
configuration
value
for
whether
it
should
be
turned
on
or
off,
and
if
it's
on
it'll
be
infinitely
polling,
every
10
seconds
or
whatever
we
configure
it
to
and
then
maybe
later
we'll
make.
Some
or
later
we
will
likely
make
some
efficiency
improvements
so
that
it's
not
pulling
continuously.
A
But
that's
basically
it
and
what
right
now
I,
like
the
immediate
helper
I,
want
right
now
is
basically
to
run
the
the
gitlab
agent
locally.
I've
had
success,
running
the
server
side,
it's
up
here,
but
the
agent
side,
Agent
K
I'm
having
this
error,
I
haven't
looked
into
it
really
deeply
like
this
is
just
like
last
20
minutes
stuff.
So
I
I,
don't
know
what
could
happen
here.
C
So
I
followed
like
there
there's
the
instructions
that
we
have
in
the
the
architecture:
dots
like
remote,
Dev
architecture,
docs,
and
then
they
reference
the
setup
docs
that
are
in
Agent
K.
Does
this
mean
like
I,
know,
there's
a
host
file,
entry,
I,
don't
know
if
that's
maybe
part
of
it.
What
do
you
think
Michael.
D
A
A
A
B
D
Oh
okay,
it's
potentially
external
address,
yeah
like
its
own
URL,
basically,
and
that
is
used
later
by
other
class
instances
to
connect
to
discuss
instance,
because
you
may
have
a
crazy
network
setup
like
reverse
proxy
in
front
of
which
Castle
a
sidecar
or
something
crazy
like
that.
It's
a
separate
thing
basically.
C
B
D
B
A
D
B
C
D
C
B
C
C
D
You
you
don't,
because
you
want
to
run
it
locally,
your
modified
version.
You
see
there
are
different
use
cases
right.
There
is
a
use
case
where
someone
wants
to
use
it
with
GDK,
but
they
are
not
developing
the
agent
they
want
to,
for
example,
develop
front-end
and
use
the
agent
to
access,
kubernetes,
API
or
backend
for
gitlab.
C
D
C
D
D
Options,
no,
not
really
not
really.
This
difference
here
is
more
like
historical
than
we
initially
had
instructions
in
the
main
reports
during
the
United
instructions
in
GDK,
and
then
they
got
better
and
these
instructions
kind
of
became
out
of
date
and
also
we
kept
adding
features
that
needed
more
of
rails.
So
you
would
and
read
GDK
basic
it's
about
locally.
D
C
A
D
A
D
D
D
B
C
A
A
A
B
B
B
A
A
D
C
D
C
A
C
B
C
C
A
C
C
C
C
C
A
C
D
D
C
D
B
A
C
C
A
C
D
A
C
A
C
A
C
D
A
D
A
I
mean
we
could
probably
rename
it
and
it
it
seems
like
it's
like
githubs,
has
two
different
polling
configurations:
yeah.
We
can
leave
those
names
for
now,
but
I
mean
we're
not
installing
anything.
It
could
be
like
polling
for
work.
You
need
to
back
off
calling
for
work
Max
back
off
or
just
if
it's
just
one
type
of
polling,
we
can
just
remove
the
the
the
prefix.
D
C
C
B
D
C
D
B
C
D
C
B
D
B
C
A
D
C
D
A
A
A
D
A
A
C
D
A
D
B
D
With
just
one
request
board,
you
will
Marshall
it
into
a
byte
slice
suggestions:
send
it
yeah
to
send
it
to
the
server
it
will
send,
set
content
type
and
set
accept.
If
you
well
actually
to
accept
a
reply,
you
need
to
provide
the
response
Handler.
You
can
see
an
example
of
that
in
the
in
the
code
you
copied
this
from
in
the
file
before
when
you
copy
this
from
here.
Above,
maybe
not
here.
D
A
C
C
A
C
C
C
C
D
C
A
A
D
C
D
A
C
So
here
is
like
the
API
endpoints,
like
what
I
threw
together
as
as
a
starting
point,
and
we
could
put
whatever
we
want
in
there
like
to
start
with.
I
just
had
work,
requests,
I.
Think
it'll
be
like
the
agent
ID
that
I
don't
even
know.
It's
gonna
be
a
project
ID.
Maybe
this
doesn't
have
to
be
tied
to
a
project
right,
but
the
agent
ID
that
works
base
ID
and
the.
C
C
C
C
C
C
C
D
C
D
A
A
D
So
agent
sends
the
agent
token
to
cast
to
authenticate
Cass
uses
JWT
secret,
but
also
it
sends
that
talking
to
rails.
So
using
that
token
rails
looks
it
up
in
the
database
and
then
it
has
the
entity
of
the
agent
and
that
has
the
ability
of
the
agent.
Obviously,
so
that's
how
you
get
the
agent
ID
that
can
be
spoofed.
A
D
Well,
I,
don't
understand,
like
agent
starts,
it
has
the
URL
and
the
token
it
doesn't
know
its
own
agent
ID.
It
needs
to
get
the
configuration.
It's
called
this
API
and
that
that
API
Returns
the
agent
ID
to
the
agent.
Then
it
uses
it
for
logging
and
blah
blah
blah.
It
never
needs
to
send
it
back
to
rares.
So.
C
B
D
Is
that
there
are
two
apis,
one
is
get
configuration,
that's
the
API
of
cast
and
that's
what
I
was
talking
about,
and
this
is
a
separate
thing.
This
is
where
it
cast
posts,
agents
configuration
too.
This
is
not
called
by
the
agent.
This
is
called
by
cast
only
and
that's
where
cast
needs
to
pass
the
ID
of
the
agent
for
which
it
also
sends
the
configuration.
So
here
it
is
the
right
thing
to
do
because
cast
can
call
the
CPI
with
any
agent
ID
for
which
it
has
the
configuration.
We
trust
us.
D
D
C
D
C
B
D
C
C
C
C
But
anyway,
like
this
is
exactly
what
I
was
saying
like
I
minimally
documented
it,
we
get
the
timestamp
at
the
most
recent
event,
so
like
we're
going
to
send
an
array
abstracts,
which
is
the
workspace
application,
identifier
and
an
array
of
the
events
and
then
the
timestamp
of
the
most
recent
event
right
in
there
that
is
sent
and
then,
if
rails,
like
applied
those
successfully
yep
everything's
done.
Maybe
I
sent
some
work
back.
Maybe
I
didn't,
but
it
echoes
back
that
time
stamp
and
then
this
is
going
to
be
in
memory.
C
C
B
A
C
Yep
we're
getting
there
I
think
just
like,
as
of
Friday,
and
yesterday,
we've
like
formalized
this
polling
architecture
and
with
mikhail's
help,
so
we
just
need
to
flesh
it
out
generate
on
the
structures
the
one
thought
I
had
is
like.
Yes,
we
need
to
hook
this
up
to
kubernetes
in
the
network
space
operator,
but
maybe
just
like
have
a
you
know:
a
mock,
a
fake
implementation
of
it
that
just
like
logs
to
the
console.
C
C
C
A
A
C
A
Least,
at
this
point,
eventually,
it
could
be
a
VM,
but
at
least
at
this
point
it's
a
Docker
container
that
contains
contains
a
remote
development
environment
and
then
maybe
just
to
hedge,
our
documentation
a
little
bit.
If
you
see
the
word
workstation,
it
is
mistaken
and
it's
supposed
to
be
workspace.
Yeah.
C
A
C
A
I
think
the
biggest
blockers
are
the
lack
of
definition,
which
we
are
like
constantly
working
on
every
every
one
of
these
meetings.
We
drill
a
bit
a
little
bit
deeper
on
on
what
we
need
to
be
doing
right,
I
think
we'll
be
firmer
on
whether
we
deliver
it
at
the
point
where
Charter
was
talking
about
where,
like
we
know
exactly
what
to
do,
and
we
can
clearly
parallelize
all
of
the
work.
C
B
D
A
C
D
D
A
Just
with
your
acknowledgments
yeah,
that's
why
we're
not
getting
work
that
that's!
That
would
be
like
the
second
iteration
we're
not
getting
work.
We're
just
we're.
Take
like
technically
we're
pushing
and
pulling
in
the
same
HTTP
request,
yeah
we're
pushing
the
events
and
pulling
any
work
as
as
well
as
the
acknowledgments
of
the
events
we
sent.