►
Description
wasmCloud is a platform for writing portable business logic that can run anywhere from the edge to the cloud, that boasts a secure-by-default, boilerplate-free developer experience with rapid feedback loop.
https://wasmcloud.com
A
Welcome
to
wasm
cloud
wednesday
for
march
the
30th
2022
we've
got
a
special
demo
today
with
ninjas.
You
just
want
to
get
us
kicked
off.
B
Yeah,
thank
you
so
last
week
I
showed
just
a
still
image
of
ninjas,
but
now
I
was
going
to
try
to
do
a
live
demo
hell
here
so
I'll
I'll.
Kick
it
off
and.
C
B
B
I
mentioned
well
last
week
the
the
parts
included
in
this
demo,
so
the
parts
that
are
running
in
using
wasm
is
the
pool
simulation
written
in
in
c
plus,
plus
and
also
the
lawrence
attractor
simulation
here,
hovering
over
the
the
pool.
B
And
then
we
have
a
lot
of
other
things
going
on,
at
the
same
time,
of
course,
very
obvious
that
the
ninjas
they're
taken
from
miximo
and
the
animations
of
of
the
mannequins
as
well.
Then
in
the
here.
In
the
background,
you
have
a
spectrum
analyzer
of
the
the
audio
being
played,
and
that
is
run
using
the
web
audio
api
in
the
browser
and
then
the
3d
rendering
itself
is
done
with
webgl.
B
And
then
also
you
have
the
possibility,
if
you
have
vr
equipment
to
go
into
vr
mode
to
get
an
a
an
immersive
experience
of
this
so
well,
I
I
I
have
the
sound.
It's
really.
A
B
My
my
earphone
so
I'll
just
turn
the
ninjas
off
for
a
minute
and
well.
If
you
have
some
some
particular
questions
about
the
demo.
So
just
please
go
ahead
and
ask.
A
I
think
would
be
great
to
hear.
I
know
the
you
shared
that
the
like
the
waves
were
written
in
c
plus,
and
things
like
that.
Yeah.
B
A
Parts
are
you
know,
what
was
it
like?
You
know
getting
it
all
ported
via
webassembly
and
getting
it
all
running.
B
So
the
the
lawrence
tractor
is
written,
rust
and
and
of
course,
since
rust
has
really
made
a
point
of
being,
I
mean
webassembly
friendly
language.
That
was
really
a
breeze
and
I
mean
when
you
compile
with
I.
I
think
it's
wasn't
pack
right.
That's
really
a
great
experience.
B
I
mean
you,
you
get
the
the
type
script
interface
is
generated
for
you
and
you
don't
have
to
do
any
sort
of
type
conversion
yourself
and
regarding
c
plus
plus
that's
a
little
bit
bumpier
for
that
you're
you're,
using
m
scriptin
and
well
you
you
got
a
a
bunch
of
canned
javascript
being
generated,
and
then
it's
really
not
that
friendly,
but
it
works
it's
so
it's
okay
and-
and
you
have
some
macros
and
I
mean
some
help,
generating
the
the
translation
between
javascript
and
and
c
plus,
plus
orgasm.
A
That
is
phenomenal
and
thank
you
again
for
sharing
that
with
us.
I'm
gonna
go
ahead
and
move
on
to
a
second
quick
talk
through
about
an
update
on
the
machine
learning
framework
for
wasmcloud
that
bmw
intel
and
our
very
own
steve
have
been
working
on
steve.
Would
it
be
you
guys
just
want
steve
that
both
ways
you
can
have
them
that
way?
A
Do
you
just
want
to
right
on
my
auto
there
steve
would
that
work
and
I'll
share
the
screen
for
you.
C
Hi
so
christoph
brewing
submitted
an
update
to
the
ml
inference
capability
provider,
and
previously
it
had
a
couple
of
really
simple
models:
it
just
did
an
identity
transformation
and
then
a
n
plus
three
transformation
just
really
to
demonstrate
that
you
could
do
an
api
and
invoke
a
predict
function
on
the
provider.
C
But
there
are
a
couple
of
updates
that
have
been
been
submitted
in
the
last
week
and
a
half
one
of
them
is
to
add
a
image
processing
stage
so
that
the
predict
function
just
takes
a
tensor
which
is
a
a
multi-dimensional
array
and
it
could
apply
to
any
kind
of
machine
learning
model.
C
One
thing
kristoff
did:
is
he
added
tensorflow
support,
so
it
now
supports
tensorflow
and
onyx,
there's
a
another
popular
framework
openvino
that
will
be
added
later
and,
in
addition,
there's
some
there's
quite
a
few
image
processing
models,
including
mobilenet,
and
so
for
the
image
preprocessing
step.
It
takes
a
image
and
transforms
it
into
that
tensor
to
hand
to
the
preprocessing
step.
That's
done
by
a
wasn't
cloud
actor.
C
Another
thing
that
was
done
is
we
updated
the
interface
for
the
inference
engine
that
predict
api
and
how
we
express
tensors,
and
some
of
that
is
based
on
some
work.
We
did
with
the
with
the
wessen
end
team,
to
up
update
the
the
way
that
we
format
the
the
arrays
and
adding
a
little
bit
of
metadata,
so
that
the
sender
and
receiver
can
agree
on
the
format
of
the
data.
So
that's
in
kristoff's.
C
Repository
that
we've
been.
A
C
There
and
I
encourage
you
to
play
with
it
and
send
us
some
feedback,
and
let
us
know
what
you
think.
A
Yeah,
it's
linked
in
the
machine
learning
channel
if
you're
on
our
slack
and
it's
also
linked
in
a
general
I
cross
posted
there.
Steve
tell
me
a
little
more.
I
think
that
there
was
a
great
discussion
around
the
idea
of
having
the
pre-processor
actors
and
what
are
the
pros
and
cons
of
that.
You
know
all
of
these
machine
learning
algorithms
that
ship
can
need
this
pre-processing
step
to
prepare
the
data.
A
So
what
was
it
about
the
kind
of
like
you
know
the
scalability
or
what
were
the
sort
of
design
considerations
that
led
you
guys
just
to
do
that
in
watson,
cloud.
C
Well
sure,
so
for
for
wasmcloud,
one
of
the
one
of
the
features
we
really
like
is
the
ability
to
find
a
provider
that
has
the
capabilities.
You
need
the
use
case
that
we're
thinking,
one
of
the
use
cases
we're
thinking
of
is
that
you,
you
might
want
to
have
the
data
processed
locally,
maybe
on
the
device
where
say,
you've
captured
an
image
or
maybe
on
a
lan
on
your
homeland
and
a
reason
for
doing
that
is
to
maintain
privacy.
So
you
don't
have
to
send
your
pictures
to
the
cloud.
C
So
there's
a
lot
of
a
lot
of
applications
where
the
we
don't
really
need
to
have
the
data
the
pixel
data
on
on
the
cloud.
C
We
just
need
to
know
the
results
like
you
took
a
picture
of
x
or
we
recognized
your
face
and
so
on,
and
so
so
wasmcloud
can
also
optionally
select,
say
a
a
server,
that's
running
a
cloud
server,
that's
running
a
high-performance,
cpu
or
or
a
bank
of
gpus,
and
if
maybe,
if
there's
not
a
local
provider,
then
it
can,
it
can
go
over
to
the
net
to
the
cloud
hosted
capability
provider
and
that's
done
through
the
magic
of
gnats,
where,
with
a
nats
leaf
node,
it
will
always
prefer
dispatching
messages
to
something
that's
on
the
local
network
or
connected
to
that
leaf,
node,
and
so
that
allows
you
to
create
that
kind
of
architecture.
C
Another
another
reason
we,
like
the
actor
model,
is
that
these
different,
these
different
machine
learning
models
all
require
data
in
a
different
format.
So
for
images
some
expect
an
rgb
format.
Some
expect
a
bgr
format.
Some
expect
the
color
data
to
be
a
color
space
from
a
number
from
zero
to
one.
Some
expect
it
to
be
in
a
range
minus
one
to
plus
one
and
that
can
vary
depending
on
whether
you're,
using
mobile
or
imagenet,
or
the
different
image
processing
models.
C
So
what
we,
what
we
want
to
be
able
to
do
dynamically
is
send
it
to
a
preprocessor,
that's
connected
to
that
model,
maybe
through
a
nickname
and
then
be
able
to
at
runtime
dynamically
switch
between
models
so
that'll,
let
us
do
say
a
b
testing.
C
If,
if
we
want
to,
we
have
a
model
we
like,
but
maybe
the
machine
learning
team
has
come
up
with
another
one
they
think
is
better.
Then
we
could
direct
five
percent
of
the
traffic
to
the
new
one
and
then
test
the
relative
performance
of
those
two
processing
models.
And
if
we
like
it,
then
we
can
increase
performance.
So
this
sort
of
classic
a
b
testing.
C
There
are
other
names
for
it
in
machine
learning
community,
but
it's
a
pretty
common
use
case
and
so
being
able
to
do
that.
Switching
is,
is
really
attractive
and
there's
ways
we
can
do
that
with
an
actor
and
change
that
at
runtime.
So
what
do
you
think
are
the
next
steps.
A
C
Well,
that's
that's
a
great
question
and,
and
part
of
that
is
because
this
is
an
open
source
project
and
we
want
to
get
people
to
participate
and
use
it.
So
part
of
that
decision
about
what's
next
on
the
roadmap
will
depend
on
the
feedback
we
get.
C
Some
of
the
things
we
want
to
do
is
really
to
build
out
some
realistic
demos
that
use
these
pipelines
so
that
we
know
we're
not
just
developing
an
abstract
api
for
the
sake
of
an
api,
but
that
we
know
it's
an
api
that
works
for
some
real
world
applications,
so
we'll
be
building
some
of
those
there's
a
few
kinds
of
problem.
Domains
that
are
interesting
anomaly
detection
is
one
image.
Recognition
is
another
and
then
also
testing
testing,
some
of
the
wasmcloud
failover
capabilities.
C
A
A
Okay,
I'll
go
ahead
and
move
on
to
the
next
announcements.
Thank
you.
So
much
for
the
update,
steve
the
invites
and
schedule
for
wazen
day
eu
in
valencia,
spain
have
been
announced
I'll
drop,
a
link
in
channel
it's
going
to
be
a
phenomenal
day
of
content,
we're
going
to
start
with
a
a
keynote
from
bailey
hayes.
A
Then
we've
got
a
talk
from
steve
sanderson
about
bringing
wazom
to
dotnet
mainstream
in
his
submission.
He
did
mention.
Wasn't
cloud.
A
We've
got
an
overview
from
divya
about
the
binary
magic
of
quasm,
the
second
state
folks,
michael
a
frequent
contributor
and
speaker
talking
about
javascript
python
and
ruby
in
webassembly,
a
presentation
from
jeff
your
colleague
over
at
shopify.
Then
then
a
lightning
talk
about
a
wasm
cloud
and
bevy
ecs,
and
this
is
one
I'm
super
interested
in
allen's
been
building.
You
know
games
and
gaming
platforms
for
a
while
now
using
wasmcloud,
and
I'm
really
curious
to
hear
his
presentation
and
dive
in
a
bit
further.
A
Then
we've
got
flavio
and
rafael
talking
about
kube
warden
off
the
susie
team,
some
of
the
admission
controllers
for
kubernetes,
an
update
on
wazi
from
nathaniel
in
harold,
then
another
lightning
talk
around
ron
evans,
who
is
the
sort
of
big
facilitator
of
the
tiny
go
community
if
you've
ever
been
on
the
tiny
ghost
slack
he's,
probably
the
gentleman
that
helped
you
out
lightning
talk
about
whammer,
which
I'm
looking
forward
to
web
assembly
on
small
devices
and
then
the
infineon
folks
are
working
on
a
really
great
use
case
for
web
assembly.
A
They've
integrated
webassembly
into
their
streaming
platform,
so
think
if
kafka
had
the
ability
to
do
in-line
transforms
on
data
and
things
like
that,
a
great
use
case
and
then
a
colin
murphy
from
adobe
we'll
be
talking
about
some
of
the
things
that
adobe
is
doing
on
web
assembly.
I'm
excited
about
this
one.
They
plan
to
talk
a
little
bit
about
what
they've
been
doing
with
laws
and
cloud
as
well,
and
then
matt
butcher,
radu
and
radu
are
going
to
talk
about
building
a
cms.
A
So
I
hope
that
I
can
meet
many
of
you
there
or
see
many
of
you
there
this
year
and
if
you
can't
all
online
tickets
to
kubecon
this
year
include
full
access
to
all
of
the
days.
So
there's
no
more
day
only
tickets
and
you'll
have
the
ability
to
sort
of
stream
all
of
the
wasom
day.
Programming
live.
If
you
can't
make
it.
A
If
you
can,
let
us
know
a
couple
of
us
will
be
over
there
this
year
and
we'll
let
you
know
some
more
about
what
we
plan
to
do
as
a
team
on
site
open
call
to
the
community.
Are
there
any
topics
to
discuss
today.
A
Anything
anybody
have
anything
that
they
would
like
to
share
across
across
or
any
community
events
brooks
any
community
call-outs.
Today,
okay,
we're
all
heads
down
this
week
on
planning
at
our
off-site
in
utah.
So
we've
we've
kind
of
got
our
hands
full
this
week,
but
I
guess
open
floor.