Structuring Inputs & Outputs in Multi Agent systems Using CrewAI

Example of Pydantic Models

Let’s make a UserModel that inherits from BaseModel from Pydantic and the instantiated class should have an “id” as int, name as “str” and email as “email” 

from pydantic import BaseModel
class UserModel(BaseModel):
   id: int
   name: str
   email: str
# Valid input
valid_user = UserModel(id=1, name="Vidhya", email="[email protected]")
print(valid_user)

id=1 name="Vidhya" email="[email protected]"

# Invalid input (raises a validation error)
try:
   invalid_user = UserModel(id="one", name="Vidhya", email="[email protected]")
except ValueError as e:
   print(f"Validation Error: {e}")

Validation Error: 1 validation error for UserModel
id
  Input should be a valid integer, unable to parse string as an integer
[type=int_parsing, input_value="one", input_type=str]
    For further information visit https://errors.pydantic.dev/2.9/v/int_parsing

We got a validation error when we tried passing a string in id and Pydantic also returns the link to documentation that helps us understand the error.

Now, let’s look at other features of Pydantic, such as optional, date, and default value. 

from pydantic import BaseModel
from typing import Optional
from datetime import date
class EventModel(BaseModel):
   event_name: Optional[str] = None  # It's Optional to provide a value
   event_loc: str = "India" # Default value
   event_date: date
# Automatic conversion of a string to a `date` object
event = EventModel(event_date="2024-10-17")
print(event)

event_name=None event_loc="India" event_date=datetime.date(2024, 10, 17)

We have passed only the event_date, yet it’s not throwing any error because we set event_name to Optional and event_loc to “India” by default. Also, notice that the date is being typecast from string to datetime. 

Installations

We will be using CrewAI throughout this guide, so make sure you install CrewAI.

!pip install crewai

Structuring Inputs in Agentic Systems

The inputs can be formatted while defining the Agent and Task using curly braces { } with a variable name inside. Setting the human_input=True will prompt the user for feedback for the output. Let’s define an agent and task to answer questions related to physics:

Note: I’ll be using the  ‘gpt-4o-mini-2024-07-18’ from openai throughout the guide. 

from crewai import Agent, Task, Crew
import os
os.environ['OPENAI_API_KEY']=''
llm_config={'model':'gpt-4o-mini-2024-07-18'}
teacher_agent = Agent(
 role="Physics Professor",
 goal="You have strong foundational knowledge on physics "
       "You give the best answer to your students",
 backstory=(
   "You work as a college professor"
   "You teach physics and clarify doubts {doubts}"
   "You answer the question only if it is related to physics"
   "You make sure that you provide the best answers!"
 ),
 allow_delegation=False,
 verbose=True)
doubt_clarification_task = Task(
   description=(
       "Address the physics doubts raised by the student {doubts}. "
       "You don't answer the question if the question is not related to physics"
       "Provide clear, detailed explanations, examples, and any necessary formulas to help the student understand the concepts."
   ),
   expected_output=(
       "A clear response to the student's doubts, "
       "including explanations and examples if needed. "
   ),
   agent=teacher_agent,
   human_input=True
)
crew = Crew(
 agents=[teacher_agent],
 tasks=[doubt_clarification_task],
 verbose=True,
 memory=False
)

The inputs can be passed to the “inputs “ parameter in crew.kickoff()

inputs = {
   "doubts": "What is thermodynamics",
}
result = crew.kickoff(inputs=inputs)

Output

# Agent: Physics Professor
## Task: Address the physics doubts raised by the student What is
thermodynamics. You don't answer the question if the question is not related
to physicsProvide clear, detailed explanations, examples, and any necessary
formulas to help the student understand the concepts.
# Agent: Physics Professor
## Final Answer: 
Thermodynamics is the branch of physics that deals with the relationships
between heat, work, temperature, and energy. It provides a framework for
understanding how energy is transferred and transformed in systems, whether
they are mechanical engines, refrigerators, or biological organisms.
The study of thermodynamics is based on four fundamental laws that describe
how energy moves and changes form:
1. **Zeroth Law of Thermodynamics**: This law establishes the concept of 
temperature. It states that if two systems are in thermal equilibrium with a
third system, then they are also in thermal equilibrium with each other. For
example, if System A is in equilibrium with System C, and System B is also
in equilibrium with System C, then System A and System B are in equilibrium 
with each other. This allows us to define temperature as a measurable entity.
2. **First Law of Thermodynamics**: Often stated as the law of energy 
conservation, this law asserts that energy cannot be created or destroyed,
only transformed. Mathematically, it can be expressed as:
   \[
   \Delta U = Q - W
   \]
   where \( \Delta U \) is the change in internal energy of the system, \( Q
\) is the heat added to the system, and \( W \) is the work done by the 
system. An example of the first law in practice is a steam engine, where heat
energy is converted into work.
3. **Second Law of Thermodynamics**: This law introduces the concept of
entropy, stating that in any energy transfer, the total entropy of an
isolated system can never decrease over time. It suggests that energy
transformations are not 100% efficient and that systems naturally progress
toward a state of disorder. A common example is the idea that heat flows
spontaneously from a hotter body to a colder body, not the reverse.
4. **Third Law of Thermodynamics**: This law states that as the temperature
of a system approaches absolute zero, the entropy of a perfect crystal
approaches zero. This implies that it is impossible to reach absolute zero
in a finite number of steps, a concept that underscores the fundamental
limits of thermodynamic processes.
In practical applications, thermodynamics is used in numerous fields such as
engineering (designing engines, refrigerators), chemistry (reaction
spontaneity), and environmental science (understanding climate systems). 
In summary, thermodynamics is essential for analyzing how energy is utilized
and transferred, providing insights into improving efficiency and
understanding natural processes in various physical systems.
 ## Final Result: Thermodynamics is the branch of physics that deals with the
relationships between heat, work, temperature, and energy. It provides a
framework for understanding how energy is transferred and transformed in
systems, whether they are mechanical engines, refrigerators, or biological
organisms.
The study of thermodynamics is based on four fundamental laws that describe
how energy moves and changes form:
1. **Zeroth Law of Thermodynamics**: This law establishes the concept of
temperature. It states that if two systems are in thermal equilibrium with a
third system, then they are also in thermal equilibrium with each other. For
example, if System A is in equilibrium with System C, and System B is also
in equilibrium with System C, then System A and System B are in equilibrium
with each other. This allows us to define temperature as a measurable
entity.
2. **First Law of Thermodynamics**: Often stated as the law of energy
conservation, this law asserts that energy cannot be created or destroyed,
only transformed. Mathematically, it can be expressed as:
   \[
   \Delta U = Q - W
   \]
   where \( \Delta U \) is the change in internal energy of the system, \( Q
\) is the heat added to the system, and \( W \) is the work done by the
system. An example of the first law in practice is a steam engine, where
heat energy is converted into work.
3. **Second Law of Thermodynamics**: This law introduces the concept of
entropy, stating that in any energy transfer, the total entropy of an
isolated system can never decrease over time. It suggests that energy
transformations are not 100% efficient and that systems naturally progress
toward a state of disorder. A common example is the idea that heat flows
spontaneously from a hotter body to a colder body, not the reverse.
4. **Third Law of Thermodynamics**: This law states that as the temperature
of a system approaches absolute zero, the entropy of a perfect crystal
approaches zero. This implies that it is impossible to reach absolute zero
in a finite number of steps, a concept that underscores the fundamental
limits of thermodynamic processes.
In practical applications, thermodynamics is used in numerous fields such as
engineering (designing engines, refrigerators), chemistry (reaction
spontaneity), and environmental science (understanding climate systems). 
In summary, thermodynamics is essential for analyzing how energy is utilized
and transferred, providing insights into improving efficiency and 
understanding natural processes in various physical systems.
=====
## Please provide feedback on the Final Result and the Agent's actions:
Give the answer in 3 lines
# Agent: Physics Professor
## Final Answer: 
Thermodynamics is the branch of physics that deals with the relationships
between heat, work, temperature, and energy. It is governed by four
fundamental laws that describe how energy is transferred and converted in
physical systems. For example, the first law of thermodynamics states that
energy cannot be created or destroyed, only transformed, which is expressed
mathematically as ΔU = Q - W, where ΔU is the change in internal energy, Q
is the heat added to the system, and W is the work done by the system.

We passed a question in the input, “What is thermodynamics?” and with the help of a human input, we got the desired answer. 

Structuring Outputs in Agentic Systems

Let’s make agents who can help me fill out details like name, email, phone number, and job one by one. Structuring the outputs as Pydantic or json helps define what output we’re expecting and the format we expect so that the subsequent agents get structured data and context.

Note: In a case like mine, not structuring the outputs with expected fields might lead to a loss of information, and the subsequent agents might start hallucinating. 

from crewai import Agent, Task, Crew
import os
os.environ['OPENAI_API_KEY']=''
llm_config={'model':'gpt-4o-mini-2024-07-18'}

Let’s define the classes using pedantic with the fields I expect in the output. As explained in the Pydantic models’ section, you can use Optional or set a value by default if needed.

from pydantic import BaseModel
from typing import List
# Define the classes using Pydantic
class nameEmail(BaseModel):
   name: List[str]
   email: List[str]
class phoneNo(BaseModel):
   name: str
   email: str
   phone: int
class allDetails(BaseModel):
   name: str
   email: List[str]
   phone: int
   job: str

Let’s define the first agent and task that fills out the user’s name and email. Use the “output_pydantic” parameter in Task(). I passed the nameEmail class as I won’t be using name and email as output here. 

# Task 1: Agent for filling name and email using output_pydantic
name_email_agent = Agent(
   role="Data Entry Specialist",
   goal="Your task is to fill out the user's name and email on your own.",
   backstory=(
       "You specialize in filling out personal information like name and email."
       "You fill the name and email on your own"
       "You use rare names and use birth year in the email"
   ),
   allow_delegation=False,
   verbose=True
)
name_email_task = Task(
   description="Fill out the name and email of the user.",
   expected_output="Name and email.",
   agent=name_email_agent,
   output_pydantic=nameEmail,
   human_input=False
)

Let’s use the other way to structure the output by using the “output_json” parameter in Task() which gives the output in json format. 

# Task 2: Agent for filling phone number using output_json
phone_number_agent = Agent(
   role="Contact Information Manager",
   goal="Your task is to fill out a random 10 digit phone number on your own.",
   backstory=(
       "You are an expert in managing contact details, specifically phone numbers."
   ),
   allow_delegation=False,
   verbose=True
)
phone_number_task = Task(
   description="Fill out the user's phone number.",
   expected_output="A JSON format with the user's phone number.",
   agent=phone_number_agent,
   output_json=phoneNo,
   human_input=False
)

Here, the agent is designed to fill out the job details for the user. The final output will contain all the name, email, phone number, and job role information as pedantic and stored in output.txt using the “output_file” parameter. 

# Task 3: Agent for filling job description using output_file
job_description_agent = Agent(
   role="Job Decider",
   goal="Your task is to randomly decide a job for the user",
   backstory=(
       "You handle assigning job roles randomly."
   ),
   allow_delegation=False,
   verbose=True
)

job_description_task = Task(

   description="Fill out the user's job description.",

   expected_output="Display name, email, phone no and job description.",

   agent=job_description_agent,

   output_pydantic=allDetails,

   output_file="/content/output.txt",

   human_input=False

)
# Define the crew to run all tasks
crew = Crew(
   agents=[name_email_agent, phone_number_agent, job_description_agent],
   tasks=[name_email_task, phone_number_task, job_description_task],
   verbose=True,
   memory=False
)
result = crew.kickoff()

# Agent: Data Entry Specialist
## Task: Fill out the name and email of the user.
# Agent: Data Entry Specialist
## Final Answer: 
Name: Elowen Thorne  
Email: [email protected]
# Agent: Contact Information Manager
## Task: Fill out the user's phone number.
# Agent: Contact Information Manager
## Final Answer: 
{
  "name": "Elowen Thorne",
  "email": "[email protected]",
  "phone_number": "1234567890"
}
# Agent: Job Decider
## Task: Fill out the user's job role.
# Agent: Job Decider
## Final Answer: 
Name: Elowen Thorne  
Email: [email protected]  
Phone No: 1234567890  
Job Role: Data Analyst

This is the output file that was created :

You can also look at each task’s output:

job_description_task.output

TaskOutput(description="Fill out the user's job role.", name=None,
expected_output="Display name, email, phone no and job role.", summary="Fill
out the user's job role....", raw='Name: Elowen Thorne  \nEmail: 
[email protected]  \nPhone No: 1234567890  \nJob Role: Data
Analyst', pydantic=allDetails(name="Elowen Thorne", email=
['[email protected]'], phone=1234567890, job='Data Analyst'),
json_dict=None, agent="Job Decider", output_format= 'pydantic'>)

I got the expected output, and due to structuring the intermediate outputs, the details remained constant as each task progressed. 

Also, to understand the Agent AI better, explore: The Agentic AI Pioneer Program.

Conclusion

This article explored the importance of structuring inputs and outputs in multi-agent using tools like Pydantic models and CrewAI. By organizing data effectively and validating inputs, we can significantly enhance the performance and reliability of multi-agent systems. Structured outputs help maintain consistency and prevent errors such as data loss and hallucinations, ensuring that agents can collaborate efficiently. Implementing these strategies allows developers to create more robust agentic systems capable of handling complex tasks with greater precision and reliability.

Frequently Asked Questions

We use cookies essential for this site to function well. Please click to help us improve its usefulness with additional cookies. Learn about our use of cookies in our Privacy Policy & Cookies Policy.

Show details

Accept all cookies

Use necessary cookies

Cookies

This site uses cookies to ensure that you get the best experience possible. To learn more about how we use cookies, please refer to our Privacy Policy & Cookies Policy.

brahmaid

It is needed for personalizing the website.

csrftoken

This cookie is used to prevent Cross-site request forgery (often abbreviated as CSRF) attacks of the website

Identityid

Preserves the login/logout state of users across the whole site.

sessionid

Preserves users’ states across page requests.

g_state

Google One-Tap login adds this g_state cookie to set the user status on how they interact with the One-Tap modal.

MUID

Used by Microsoft Clarity, to store and track visits across websites.

_clck

Used by Microsoft Clarity, Persists the Clarity User ID and preferences, unique to that site, on the browser. This ensures that behavior in subsequent visits to the same site will be attributed to the same user ID.

_clsk

Used by Microsoft Clarity, Connects multiple page views by a user into a single Clarity session recording.

SRM_I

Collects user data is specifically adapted to the user or device. The user can also be followed outside of the loaded website, creating a picture of the visitor’s behavior.

SM

Use to measure the use of the website for internal analytics

CLID

The cookie is set by embedded Microsoft Clarity scripts. The purpose of this cookie is for heatmap and session recording.

SRM_B

Collected user data is specifically adapted to the user or device. The user can also be followed outside of the loaded website, creating a picture of the visitor’s behavior.

_gid

This cookie is installed by Google Analytics. The cookie is used to store information of how visitors use a website and helps in creating an analytics report of how the website is doing. The data collected includes the number of visitors, the source where they have come from, and the pages visited in an anonymous form.

_ga_#

Used by Google Analytics, to store and count pageviews.

_gat_#

Used by Google Analytics to collect data on the number of times a user has visited the website as well as dates for the first and most recent visit.

collect

Used to send data to Google Analytics about the visitor’s device and behavior. Tracks the visitor across devices and marketing channels.

AEC

cookies ensure that requests within a browsing session are made by the user, and not by other sites.

G_ENABLED_IDPS

use the cookie when customers want to make a referral from their gmail contacts; it helps auth the gmail account.

test_cookie

This cookie is set by DoubleClick (which is owned by Google) to determine if the website visitor’s browser supports cookies.

_we_us

this is used to send push notification using webengage.

WebKlipperAuth

used by webenage to track auth of webenagage.

ln_or

Linkedin sets this cookie to registers statistical data on users’ behavior on the website for internal analytics.

JSESSIONID

Use to maintain an anonymous user session by the server.

li_rm

Used as part of the LinkedIn Remember Me feature and is set when a user clicks Remember Me on the device to make it easier for him or her to sign in to that device.

AnalyticsSyncHistory

Used to store information about the time a sync with the lms_analytics cookie took place for users in the Designated Countries.

lms_analytics

Used to store information about the time a sync with the AnalyticsSyncHistory cookie took place for users in the Designated Countries.

liap

Cookie used for Sign-in with Linkedin and/or to allow for the Linkedin follow feature.

visit

allow for the Linkedin follow feature.

li_at

often used to identify you, including your name, interests, and previous activity.

s_plt

Tracks the time that the previous page took to load

lang

Used to remember a user’s language setting to ensure LinkedIn.com displays in the language selected by the user in their settings

s_tp

Tracks percent of page viewed

AMCV_14215E3D5995C57C0A495C55%40AdobeOrg

Indicates the start of a session for Adobe Experience Cloud

s_pltp

Provides page name value (URL) for use by Adobe Analytics

s_tslv

Used to retain and fetch time since last visit in Adobe Analytics

li_theme

Remembers a user’s display preference/theme setting

li_theme_set

Remembers which users have updated their display / theme preferences

We do not use cookies of this type.

_gcl_au

Used by Google Adsense, to store and track conversions.

SID

Save certain preferences, for example the number of search results per page or activation of the SafeSearch Filter. Adjusts the ads that appear in Google Search.

SAPISID

Save certain preferences, for example the number of search results per page or activation of the SafeSearch Filter. Adjusts the ads that appear in Google Search.

__Secure-#

Save certain preferences, for example the number of search results per page or activation of the SafeSearch Filter. Adjusts the ads that appear in Google Search.

APISID

Save certain preferences, for example the number of search results per page or activation of the SafeSearch Filter. Adjusts the ads that appear in Google Search.

SSID

Save certain preferences, for example the number of search results per page or activation of the SafeSearch Filter. Adjusts the ads that appear in Google Search.

HSID

Save certain preferences, for example the number of search results per page or activation of the SafeSearch Filter. Adjusts the ads that appear in Google Search.

DV

These cookies are used for the purpose of targeted advertising.

NID

These cookies are used for the purpose of targeted advertising.

1P_JAR

These cookies are used to gather website statistics, and track conversion rates.

OTZ

Aggregate analysis of website visitors

_fbp

This cookie is set by Facebook to deliver advertisements when they are on Facebook or a digital platform powered by Facebook advertising after visiting this website.

fr

Contains a unique browser and user ID, used for targeted advertising.

bscookie

Used by LinkedIn to track the use of embedded services.

lidc

Used by LinkedIn for tracking the use of embedded services.

bcookie

Used by LinkedIn to track the use of embedded services.

aam_uuid

Use these cookies to assign a unique ID when users visit a website.

UserMatchHistory

These cookies are set by LinkedIn for advertising purposes, including: tracking visitors so that more relevant ads can be presented, allowing users to use the ‘Apply with LinkedIn’ or the ‘Sign-in with LinkedIn’ functions, collecting information about how visitors use the site, etc.

li_sugr

Used to make a probabilistic match of a user’s identity outside the Designated Countries

MR

Used to collect information for analytics purposes.

ANONCHK

Used to store session ID for a users session to ensure that clicks from adverts on the Bing search engine are verified for reporting purposes and for personalisation

We do not use cookies of this type.

Cookie declaration last updated on 24/03/2023 by Analytics Vidhya.

Cookies are small text files that can be used by websites to make a user’s experience more efficient. The law states that we can store cookies on your device if they are strictly necessary for the operation of this site. For all other types of cookies, we need your permission. This site uses different types of cookies. Some cookies are placed by third-party services that appear on our pages. Learn more about who we are, how you can contact us, and how we process personal data in our Privacy Policy.

GenAI
Pinnacle
Program

Revolutionizing AI Learning & Development

• 1:1 Mentorship with Generative AI experts
• Advanced Curriculum with 200+ Hours of Learning
• Master 26+ GenAI Tools and Libraries

Enter email address to continue

Resend OTP

Resend OTP in 45s