AI agent frameworks

사진: Unsplash 의 Ben Lambert

이전글 - AI agent, CoT, ReAct에 대한 설명

2024.11.15 - [DEV] - AI Agent concept

AI Agent concept

 

AI agent frameworks

다양한 AI agent framework들이 존재

LangChain(LangGraph), LLamaIndex, AutoGen, crewAI, Transformers Agent....

crewAI

https://www.crewai.com/

crewAI에 대해 학습 가능한 deeplearning AI short course

https://www.deeplearning.ai/short-courses/multi-ai-agent-systems-with-crewai/

Multi AI Agent Systems with crewAI

CrewAI의 에이전트는 LangChain으로 코딩된 ReAct 에이전트

crewAI reAct

 

작업을 완료했다고 생각할 때까지 루프에서 CoT 프롬프트를 수행.

Thought, Action, Observation 할 때마다 = money

아무튼 개발자가 해야 하는 일은 Agent를 만들고, agent가 수행할 Task 생성, agent가 사용할 수 있는 Tool을 생성, Process를 생성

crewAI의 높은 추상화위에 위 component들을 작업하지만 내부적으론 reAct 기반으로 동작

 

crewAI 구조

 

Task

 - agent가 수행할 작업, 각 작업은 agent에게 할당

Agent

 - 특정업무를 수행할 AI agent, 각 agent는 다른 작업의 전문가

Tool

 - agent가 작업을 수행할 때 사용하는 도구

 - 검색엔진, 번역, 계산...

process

 - 에이전트가 함께 작업하는 방식지시

 - 순차적, 병렬적, 관리자 agent가 일반 agent를 할당

Example

여행 계획 agent

https://github.com/crewAIInc/crewAI-examples/tree/main/trip_planner

 

여행계획 agent 생성

class TripAgents():

  def city_selection_agent(self):
    return Agent(
        role='City Selection Expert',
        goal='Select the best city based on weather, season, and prices',
        backstory=
        'An expert in analyzing travel data to pick ideal destinations',
        tools=[
            SearchTools.search_internet,
            BrowserTools.scrape_and_summarize_website,
        ],
        verbose=True)

  def local_expert(self):
    return Agent(
        role='Local Expert at this city',
        goal='Provide the BEST insights about the selected city',
        backstory="""A knowledgeable local guide with extensive information
        about the city, it's attractions and customs""",
        tools=[
            SearchTools.search_internet,
            BrowserTools.scrape_and_summarize_website,
        ],
        verbose=True)

  def travel_concierge(self):
    return Agent(
        role='Amazing Travel Concierge',
        goal="""Create the most amazing travel itineraries with budget and 
        packing suggestions for the city""",
        backstory="""Specialist in travel planning and logistics with 
        decades of experience""",
        tools=[
            SearchTools.search_internet,
            BrowserTools.scrape_and_summarize_website,
            CalculatorTools.calculate,
        ],
        verbose=True)

 

여행계획시 실행할 task설정

class TripTasks:

    def identify_task(self, agent, origin, cities, interests, range):
        return Task(
            description=dedent(f"""
                Analyze and select the best city for the trip based 
                on specific criteria such as weather patterns, seasonal
                events, and travel costs. This task involves comparing
                multiple cities, considering factors like current weather
                conditions, upcoming cultural or seasonal events, and
                overall travel expenses. 
                
                Your final answer must be a detailed
                report on the chosen city, and everything you found out
                about it, including the actual flight costs, weather 
                forecast and attractions.
                {self.__tip_section()}

                Traveling from: {origin}
                City Options: {cities}
                Trip Date: {range}
                Traveler Interests: {interests}
            """),
            agent=agent,
            expected_output="Detailed report on the chosen city including flight costs, weather forecast, and attractions"
        )

    def gather_task(self, agent, origin, interests, range):
        return Task(
            description=dedent(f"""
                As a local expert on this city you must compile an 
                in-depth guide for someone traveling there and wanting 
                to have THE BEST trip ever!
                Gather information about key attractions, local customs,
                special events, and daily activity recommendations.
                Find the best spots to go to, the kind of place only a
                local would know.
                This guide should provide a thorough overview of what 
                the city has to offer, including hidden gems, cultural
                hotspots, must-visit landmarks, weather forecasts, and
                high level costs.
                
                The final answer must be a comprehensive city guide, 
                rich in cultural insights and practical tips, 
                tailored to enhance the travel experience.
                {self.__tip_section()}

                Trip Date: {range}
                Traveling from: {origin}
                Traveler Interests: {interests}
            """),
            agent=agent,
            expected_output="Comprehensive city guide including hidden gems, cultural hotspots, and practical travel tips"
        )

    def plan_task(self, agent, origin, interests, range):
        return Task(
            description=dedent(f"""
                Expand this guide into a full 7-day travel 
                itinerary with detailed per-day plans, including 
                weather forecasts, places to eat, packing suggestions, 
                and a budget breakdown.
                
                You MUST suggest actual places to visit, actual hotels 
                to stay and actual restaurants to go to.
                
                This itinerary should cover all aspects of the trip, 
                from arrival to departure, integrating the city guide
                information with practical travel logistics.
                
                Your final answer MUST be a complete expanded travel plan,
                formatted as markdown, encompassing a daily schedule,
                anticipated weather conditions, recommended clothing and
                items to pack, and a detailed budget, ensuring THE BEST
                TRIP EVER. Be specific and give it a reason why you picked
                each place, what makes them special! {self.__tip_section()}

                Trip Date: {range}
                Traveling from: {origin}
                Traveler Interests: {interests}
            """),
            agent=agent,
            expected_output="Complete expanded travel plan with daily schedule, weather conditions, packing suggestions, and budget breakdown"
        )

 

생성한 agent, task 연결하여 crew.kickoff

agents = TripAgents()
    tasks = TripTasks()

    city_selector_agent = agents.city_selection_agent()
    local_expert_agent = agents.local_expert()
    travel_concierge_agent = agents.travel_concierge()

    identify_task = tasks.identify_task(
      city_selector_agent,
      self.origin,
      self.cities,
      self.interests,
      self.date_range
    )
    gather_task = tasks.gather_task(
      local_expert_agent,
      self.origin,
      self.interests,
      self.date_range
    )
    plan_task = tasks.plan_task(
      travel_concierge_agent, 
      self.origin,
      self.interests,
      self.date_range
    )

    crew = Crew(
      agents=[
        city_selector_agent, local_expert_agent, travel_concierge_agent
      ],
      tasks=[identify_task, gather_task, plan_task],
      verbose=True
    )

    result = crew.kickoff()
    return result

 

genspark의 sparkpages와 비슷한 결과 생성

https://www.genspark.ai/travel

genspark travel page

genspark의 sparkpages나 perplexity의 발견하기도 이런 방식으로 만들지 않았을까 싶고, 비슷하게 만들어 볼 수 있을 것 같다.
query time에 생성하면 시간, 비용이 너무 크고, batch time에 문서를 만들어 두고 제공하는 방식이 되어야 할 것 같고, 위 서비스들도 그렇게 하는 것으로 보임

genspark 우상단 'sparkpage 생성'을 클릭해보면 주제를 입력하면 실시간으로 주제 관련된 문서를 작성

 

sparkpage 생성

 

생성하는데 꽤 오래 걸림

생성된 품질 좋은 페이지의 엔티티를 지식그래프로 연결하거나 벡터디비에 넣고 생성형 검색, 일반 검색의 소스로 사용해도 좋을 듯
만약 생성된 문서를 서비스 하려면 문서의 품질을 평가 할 수 있는 기능도 필요해보인다.