This project is dedicated to automating business workflows using multi-agent AI systems. By leveraging the power of autonomous AI agents, this framework enables efficient and effective performance of complex, multi-step tasks.
Designing effective AI agents and organize team of AI agents them to perform complex, multi-step tasks.
Provide human feedback iteratively to fine-tune response
When LLMs operate autonomously, they become agents. AI Agents ask and answer questions on its own.
LLMs + Cognition = AI Agents.
Source: deeplearning.ai
Framework for building multi-agent systems (that are autonomous, role-playing and collaborate)
crew : Team of AI agents working together, each with a specific role.
- Assign specific role and specific task to each agent and improved output. Eg. One agent does exhaustive research and other does professional writing.
- Use different LLMs for specific tasks
Source: deeplearning.ai
- Resume Strategist : Tailor resumes and interview prep
- Design, build and test website
- Research, write and fact-check technical papers
- Automate customer support inquiries
- Conduct social media campaigns
- Perform financial analysis
New way to write software. Provide fizzy inputs, apply fuzzy tranformations and get fuzzy outputs.
Reason why people love chatGPT: Probablistic nature
Source: deeplearning.ai
- Capture information about the company
- Use classification to generate scores for company
- Prioritise for sales
Source: deeplearning.ai
- AI agent research about company (via Google, internal database)
- AI agent compares companies (new ones, old ones)
- AI agent scores companies (based on parameters)
- AI agent provides intelligent questions to ask based on scores
Source: deeplearning.ai
- Role: Assign specialized role to agents
- Memory: Provide agents with short-term, long-term and entity memory
- Tools: Assign pre-built and custom tools to each agent (eg. for web search)
- Focus: Break down task, goals and tools and assign multiple AI agents for better performance
- Guardrails: Effectively handle errors, hallucinations and infinite loops.
- Cooperation: Perform tasks in series, in parallel and hierarchical fashion
Source: deeplearning.ai
More specific role = Better response. Gives clear idea about agent's function in the crew.
Example: You are a financial analyst v/s you are FINRA approved financial analyst.
from crewai import Agent
agent = Agent(
role='Data Analyst',
goal='Extract actionable insights',
backstory="""You're a data analyst at a large company.
You're responsible for analyzing data and providing insights
to the business."""
)Assinging too many tasks, tools, context to a single agent, cause losing essential information and hallucinate.
Therefore, break down task, goals and tools and assign to multiple AI agents for better performance
research_ai_task = Task(
description='Find and summarize the latest AI news',
expected_output='A bullet list summary of the top 5 most important AI news',
agent=research_agent,
tools=[search_tool]
)
research_ops_task = Task(
description='Find and summarize the latest AI Ops news',
expected_output='A bullet list summary of the top 5 most important AI Ops news',
agent=research_agent,
tools=[search_tool]
)
write_blog_task = Task(
description="Write a full blog post about the importance of AI and its latest news",
expected_output='Full blog post that is 4 paragraphs long',
agent=writer_agent,
context=[research_ai_task, research_ops_task]
)Assign tools to AI Agents and Tasks for improving execution and performance.
from crewai import Agent
researcher = Agent(
role='Market Research Analyst',
goal='Provide up-to-date market analysis of the AI industry',
backstory='An expert analyst with a keen eye for market trends.',
tools=[search_tool, web_rag_tool]
)Note: Tasks specific tools override an agent's default tools.
task = Task(
description='Find and summarize the latest AI news',
expected_output='A bullet list summary of the top 5 most important AI news',
agent=research_agent,
tools=[search_tool]
)Agents collobrate to combine skills, share information, delegate tasks to each other.
Ideal for projects requiring tasks to be completed in a specific order.
report_crew = Crew(
agents=[researcher, analyst, writer],
tasks=[research_task, analysis_task, writing_task], # tasks executed in the order of listing, with output of one task serving as context for the next
process=Process.sequential
)- CrewAI automatically creates a manager agent, requiring the specification of a manager language model (manager_llm) for the manager agent.
- THe manager allocates tasks among crew members based on their roles, tools and capabilities.
- The manager evaluates outcomes to ensure they meet the required standards.
- set Process attribute to Process.hierarchical for Crew object
- set manager_llm for Crew Object. Mandatory for hierarchical process
from crewai import Crew
from crewai.process import Process
from langchain_openai import ChatOpenAI
# Example: Creating a crew with a hierarchical process
# Ensure to provide a manager_llm
crew = Crew(
agents=my_agents,
tasks=my_tasks,
process=Process.hierarchical,
manager_llm=ChatOpenAI(model="gpt-4")
)Tasks can now be executed asynchronously, allowing for parallel processing and efficiency improvements
list_ideas = Task(
description="List of 5 interesting ideas to explore for an article about AI.",
expected_output="Bullet point list of 5 ideas for an article.",
agent=researcher,
async_execution=True # Will be executed asynchronously
)
list_important_history = Task(
description="Research the history of AI and give me the 5 most important events.",
expected_output="Bullet point list of 5 important events.",
agent=researcher,
async_execution=True # Will be executed asynchronously
)
write_article = Task(
description="Write an article about AI, its history, and interesting ideas.",
expected_output="A 4 paragraph article about AI.",
agent=writer,
context=[list_ideas, list_important_history] # Will wait for the output of the two tasks to be completed
)Implemented at Framework level to prevrnt hallucinations, errors and infintite loops.
CrewAI provides short-term memory, long-term memory, entity memory, and newly identified contextual memory to help AI agents to remember, reason, and learn from past interactions.
Advantages of Memory
- More contexual awareness, leading to more coherent and relevant responses
- Experience Accumulation, learning from past actions to improve future decision-making and problem-solving.
- Entity Understanding, agents can recognize and remember key entities, enhancing understanding.
Source: deeplearning.ai
Enable memory by setting memory=True in the Crew objects arguments.
from crewai import Crew, Agent, Task, Process
# Assemble your crew with memory capabilities
my_crew = Crew(
agents=[...],
tasks=[...],
process=Process.sequential,
memory=True,
verbose=True
)
Source: deeplearning.ai
Think of yourself as a Manager
Answer 3 questions:
- What is the Goal ?
- What is the Process ?
- What kind of people I would like to hire, to get the work done
This will help to create agents (roles, goals, backstory)
Source: deeplearning.ai
- Versatile: Hndle Fuzzy inputs and provide strongly typed outputs
- Caching Mechanism: Reuse previous results. Caching layer prevent unnecessary requests, stay within rate limits, speed up execution time
- Error Handling: Gracefully handle erors & exceptions. How ? Sending error message to agent and ask agent to retry
NOTE: CrewAI supports both crewAI Toolkit and LangChain Tools
Think of yourself as a Manager
Ask what kind of process and tasks I expect individuals on my team to do.
Task requires min. 3 things:
- description
- expected_output
- agent that will perform the task
Source: deeplearning.ai
Initial context fades away as tasks flows from agent to agent.
Source: deeplearning.ai
- Manager always remeber initial goal
- Automatically delegates tasks
- Asks agents for further improvement, if required.