The Microsoft Agent Framework is an open-source engine designed to build and orchestrate intelligent AI agents, demonstrated through a workflow that automates the tedious process of evaluating clothing line pitches for a company named Zava.
Mind Map
클릭해서 펼치기
클릭해서 인터랙티브 마인드맵 전체 보기
Hi, my name is Elijah Strait and I am
super excited to be able to demo for you
today the Microsoft Agent Framework. The
Microsoft Agent Framework is a new
open-source engine for building and
orchestrating intelligent AI agents.
Today we're going to be looking at a
clothing company called Zava. Zava
receives a lot of different pitches for
various clothing lines that they want to
people want them to incorporate into
their overall product line. And this is
tedious to have to be able to go through
all of these different pitch decks for
these various types of close. And so
we're going to see how the agent
framework workflows are able to help
Zava automate this process and reduce
toil. What's going to happen is first
the pitch decks are going to be uh
brought in using a Python script in a
Python package. What's great about the
workflow uh architecture is that you're
able to mix both traditional business
processes with multi- aent solutions.
And so from there, after extracting all
of the information from the pitch deck,
that information is going to go to a
subworkflow of concurrent agents that
are going to do market research, fashion
design research, as well as production
feasibility research. The output of
those agents is then going to go to an
aggregator. From there, the a report is
going to be drafted. A human designer is
going to be able to look at the report
and see if this is a product line they
want to incorporate or not. And then a
final report is then drafted. So today
we will be actually looking at an
extremely bad pitch deck called the
Arctic Blast Coats heavy summer
collection. Now this is obviously not a
great collection. Um it's 95 degrees in
the summertime. Nobody wants coats. Not
have they don't have a very good target
customer. Um the product line is not
awesome. Sales projections are not
great. And they're really just like not
a very good company or or product. So,
let's see if our
workflow can identify the problems with
this. While that's running, we're going
to go over here to VS Code and check out
how we're combining both our traditional
business processes with our agents. And
so, as I mentioned earlier, we're just
going to extract um the the content of
these PowerPoint decks using a Python
script and a package. So, you can see
here, we're going to validate that um
the file exists as a PowerPoint file.
load said file, initialize a data
structure for getting the the
information out, process each slide,
making sure we're getting um text out
from each shape in the slide, and then
identifying some keywords. Then from
there, after we've done that, we're
going to initialize our agents. And so
you can see here, we're creating an
agent with the system prompt, your
senior fashion analysis research at Zava
or excuse me, market uh research anal
analyst at Zava. Made sure to that the
output is very concise. We don't want it
to be too verbose for us. And we're
going to be actually using the Azure AI
Foundry uh agent service in order to do
this using the project client. Um and
then you can see we're creating some
other agents here as well. And so then
how do we get these to all come
together? We're going to be using the
workflow builder. And so in workflows we
have these concepts of uh executors and
edges, which is very similar to nodes
and edges if you're familiar with
directed graphs. And we're going to
start with uh and the beginning executor
then add a bunch of edges connecting it.
So you can see here we have our
concurrent analysis subworkflow which is
our agents. We also have our report
writer for writing our report. We have a
approval manager for the human approval
step. And then you can see here we also
have edges with conditional these are
conditional edges that depending on what
the human says uh either approves or
rejects it then sends it to a one edge
or another. So we can see here if it the
um appro the concept is approved it goes
to a save approved concept report and if
it's rejected it drafts for us a nice
email um to send to the designers
letting them know like hey we're not
going to go forward with this. So,
switching back to our clothing designer,
we can see here that we've gotten to the
human in the loop step and it is uh
wanting us to it says actually, hey, the
decision we want to reject it misaligns
with our brand
and um it it heavy winter collection
poorly targets audience needs and
seasonal demand. Um, a lot of risk. So,
we're going to go ahead and reject this
concept. And then what it's going to do
for us is actually create, as I
mentioned earlier, a nice email saying,
"Hey, concept designer, thank you for
submitting your clothing concept to
Zava. We appreciate your time, but we're
going to move forward with other
projects this time." That being said, we
are super excited for you guys to be
able to try out the Microsoft agent
framework and excited to see what you
텍스트나 타임스탬프를 클릭하면 동영상의 해당 장면으로 바로 이동합니다
공유:
대부분의 자막은 5초 이내에 준비됩니다
원클릭 복사125개 이상의 언어내용 검색타임스탬프로 이동
YouTube URL 붙여넣기
YouTube 동영상 링크를 입력하면 전체 자막을 가져옵니다
자막 추출 양식
대부분의 자막은 5초 이내에 준비됩니다
Chrome 확장 프로그램 설치
YouTube를 떠나지 않고 자막을 즉시 가져오세요. Chrome 확장 프로그램을 설치하면 동영상 시청 페이지에서 바로 자막에 원클릭으로 접근할 수 있습니다.
