Initially, when ChatGPT simply appeared, we used easy prompts to get solutions to our questions. Then, we encountered points with hallucinations and commenced utilizing RAG (Retrieval Augmented Era) to supply extra context to LLMs. After that, we began experimenting with AI brokers, the place LLMs act as a reasoning engine and might resolve what to do subsequent, which instruments to make use of, and when to return the ultimate reply.
The following evolutionary step is to create groups of such brokers that may collaborate with one another. This method is logical because it mirrors human interactions. We work in groups the place every member has a selected position:
- The product supervisor proposes the following challenge to work on.
- The designer creates its feel and look.
- The software program engineer develops the answer.
- The analyst examines the information to make sure it performs as anticipated and identifies methods to enhance the product for purchasers.
Equally, we will create a crew of AI brokers, every specializing in one area. They will collaborate and attain a remaining conclusion collectively. Simply as specialization enhances efficiency in actual life, it may additionally profit the efficiency of AI brokers.
One other benefit of this method is elevated flexibility. Every agent can function with its personal immediate, set of instruments and even LLM. As an illustration, we will use completely different fashions for various components of our system. You should utilize GPT-4 for the agent that wants extra reasoning and GPT-3.5 for the one which does solely easy extraction. We will even fine-tune the mannequin for small particular duties and use it in our crew of brokers.
The potential drawbacks of this method are time and price. A number of interactions and information sharing between brokers require extra calls to LLM and devour further tokens. This might end in longer wait occasions and elevated bills.
There are a number of frameworks obtainable for multi-agent methods at this time.
Listed here are among the hottest ones:
- AutoGen: Developed by Microsoft, AutoGen makes use of a conversational method and was one of many earliest frameworks for multi-agent methods,
- LangGraph: Whereas not strictly a multi-agent framework, LangGraph permits for outlining complicated interactions between actors utilizing a graph construction. So, it will also be tailored to create multi-agent methods.
- CrewAI: Positioned as a high-level framework, CrewAI facilitates the creation of “crews” consisting of role-playing brokers able to collaborating in varied methods.
I’ve determined to begin experimenting with multi-agent frameworks from CrewAI because it’s fairly extensively fashionable and person pleasant. So, it seems to be like a superb choice to start with.
On this article, I’ll stroll you thru find out how to use CrewAI. As analysts, we’re the area consultants answerable for documenting varied information sources and addressing associated questions. We’ll discover find out how to automate these duties utilizing multi-agent frameworks.
Let’s begin with establishing the atmosphere. First, we have to set up the CrewAI major bundle and an extension to work with instruments.
pip set up crewai
pip set up 'crewai[tools]'
CrewAI was developed to work primarily with OpenAI API, however I might additionally prefer to strive it with a neighborhood mannequin. Based on the ChatBot Arena Leaderboard, the perfect mannequin you may run in your laptop computer is Llama 3 (8b parameters). It will likely be probably the most possible choice for our use case.
We will entry Llama fashions utilizing Ollama. Set up is fairly simple. You must obtain Ollama from the website after which undergo the set up course of. That’s it.
Now, you may take a look at the mannequin in CLI by working the next command.
ollama run llama3
For instance, you may ask one thing like this.
Let’s create a customized Ollama mannequin to make use of later in CrewAI.
We are going to begin with a ModelFile (documentation). I solely specified the bottom mannequin (llama3), temperature and cease sequence. Nonetheless, you would possibly add extra options. For instance, you may decide the system message utilizing SYSTEM key phrase.
FROM llama3# set parameters
PARAMETER temperature 0.5
PARAMETER cease Outcome
I’ve saved it right into a Llama3ModelFile file.
Let’s create a bash script to load the bottom mannequin for Ollama and create the customized mannequin we outlined in ModelFile.
#!/bin/zsh# outline variables
model_name="llama3"
custom_model_name="crewai-llama3"
# load the bottom mannequin
ollama pull $model_name
# create the mannequin file
ollama create $custom_model_name -f ./Llama3ModelFile
Let’s execute this file.
chmod +x ./llama3_setup.sh
./llama3_setup.sh
You will discover each information on GitHub: Llama3ModelFile and llama3_setup.sh
We have to initialise the next environmental variables to make use of the native Llama mannequin with CrewAI.
os.environ["OPENAI_API_BASE"]='http://localhost:11434/v1'os.environ["OPENAI_MODEL_NAME"]='crewai-llama3'
# custom_model_name from the bash script
os.environ["OPENAI_API_KEY"] = "NA"
We’ve completed the setup and are able to proceed our journey.
As analysts, we regularly play the position of subject material consultants for information and a few data-related instruments. In my earlier crew, we used to have a channel with nearly 1K individuals, the place we had been answering numerous questions on our information and the ClickHouse database we used as storage. It took us numerous time to handle this channel. It could be attention-grabbing to see whether or not such duties may be automated with LLMs.
For this instance, I’ll use the ClickHouse database. If you happen to’re , You possibly can study extra about ClickHouse and find out how to set it up domestically in my previous article. Nonetheless, we gained’t utilise any ClickHouse-specific options, so be at liberty to stay to the database you recognize.
I’ve created a fairly easy information mannequin to work with. There are simply two tables in our DWH (Knowledge Warehouse): ecommerce_db.customers and ecommerce_db.periods. As you would possibly guess, the primary desk accommodates details about the customers of our service.
The ecommerce_db.periods desk shops details about person periods.
Concerning information supply administration, analysts usually deal with duties like writing and updating documentation and answering questions on this information. So, we’ll use LLM to jot down documentation for the desk within the database and educate it to reply questions on information or ClickHouse.
However earlier than shifting on to the implementation, let’s study extra concerning the CrewAI framework and its core ideas.
The cornerstone of a multi-agent framework is an agent idea. In CrewAI, brokers are powered by role-playing. Position-playing is a tactic once you ask an agent to undertake a persona and behave like a top-notch backend engineer or useful buyer assist agent. So, when making a CrewAI agent, you might want to specify every agent’s position, purpose, and backstory in order that LLM is aware of sufficient to play this position.
The brokers’ capabilities are restricted with out tools (features that brokers can execute and get outcomes). With CrewAI, you should use one of many predefined instruments (for instance, to look the Web, parse an internet site, or do RAG on a doc), create a customized instrument your self or use LangChain instruments. So, it’s fairly straightforward to create a robust agent.
Let’s transfer on from brokers to the work they’re doing. Brokers are engaged on tasks (particular assignments). For every activity, we have to outline an outline, anticipated output (definition of finished), set of accessible instruments and assigned agent. I actually like that these frameworks comply with the managerial greatest practices like a transparent definition of finished for the duties.
The following query is find out how to outline the execution order for duties: which one to work on first, which of them can run in parallel, and many others. CrewAI applied processes to orchestrate the duties. It gives a few choices:
- Sequential —probably the most simple method when duties are referred to as one after one other.
- Hierarchical — when there’s a supervisor (specified as LLM mannequin) that creates and delegates duties to the brokers.
Additionally, CrewAI is engaged on a consensual course of. In such a course of, brokers will be capable to make selections collaboratively with a democratic method.
There are different levers you should use to tweak the method of duties’ execution:
- You possibly can mark duties as “asynchronous”, then they are going to be executed in parallel, so it is possible for you to to get a solution quicker.
- You should utilize the “human enter” flag on a activity, after which the agent will ask for human approval earlier than finalising the output of this activity. It may possibly mean you can add an oversight to the method.
We’ve outlined all the first constructing blocks and might focus on the holly grail of CrewAI — crew idea. The crew represents the crew of brokers and the set of duties they are going to be engaged on. The method for collaboration (processes we mentioned above) will also be outlined on the crew stage.
Additionally, we will arrange the memory for a crew. Reminiscence is essential for environment friendly collaboration between the brokers. CrewAI helps three ranges of reminiscence:
- Quick-term reminiscence shops info associated to the present execution. It helps brokers to work collectively on the present activity.
- Lengthy-term reminiscence is information concerning the earlier executions saved within the native database. The sort of reminiscence permits brokers to study from earlier iterations and enhance over time.
- Entity reminiscence captures and constructions details about entities (like personas, cities, and many others.)
Proper now, you may solely change on all kinds of reminiscence for a crew with none additional customisation. Nonetheless, it doesn’t work with the Llama fashions.
We’ve discovered sufficient concerning the CrewAI framework, so it’s time to begin utilizing this information in apply.
Let’s begin with a easy activity: placing collectively the documentation for our DWH. As we mentioned earlier than, there are two tables in our DWH, and I wish to create an in depth description for them utilizing LLMs.
First method
At first, we want to consider the crew construction. Consider this as a typical managerial activity. Who would you rent for such a job?
I might break this activity into two components: retrieving information from a database and writing documentation. So, we want a database specialist and a technical author. The database specialist wants entry to a database, whereas the author gained’t want any particular instruments.
Now, we have now a high-level plan. Let’s create the brokers.
For every agent, I’ve specified the position, purpose and backstory. I’ve tried my greatest to supply brokers with all of the wanted context.
database_specialist_agent = Agent(
position = "Database specialist",
purpose = "Present information to reply enterprise questions utilizing SQL",
backstory = '''You're an knowledgeable in SQL, so you may assist the crew
to collect wanted information to energy their selections.
You're very correct and have in mind all of the nuances in information.''',
allow_delegation = False,
verbose = True
)tech_writer_agent = Agent(
position = "Technical author",
purpose = '''Write partaking and factually correct technical documentation
for information sources or instruments''',
backstory = '''
You're an knowledgeable in each expertise and communications, so you may simply clarify even subtle ideas.
You base your work on the factual info offered by your colleagues.
Your texts are concise and may be simply understood by a large viewers.
You utilize skilled however somewhat a casual type in your communication.
''',
allow_delegation = False,
verbose = True
)
We are going to use a easy sequential course of, so there’s no want for brokers to delegate duties to one another. That’s why I specified allow_delegation = False.
The following step is setting the duties for brokers. However earlier than shifting to them, we have to create a customized instrument to connect with the database.
First, I put collectively a perform to execute ClickHouse queries utilizing HTTP API.
CH_HOST = 'http://localhost:8123' # default handle def get_clickhouse_data(question, host = CH_HOST, connection_timeout = 1500):
r = requests.publish(host, params = {'question': question},
timeout = connection_timeout)
if r.status_code == 200:
return r.textual content
else:
return 'Database returned the next error:n' + r.textual content
When working with LLM brokers, it’s essential to make instruments fault-tolerant. For instance, if the database returns an error (status_code != 200), my code gained’t throw an exception. As a substitute, it’ll return the error description to the LLM so it may try to resolve the difficulty.
To create a CrewAI customized instrument, we have to derive our class from crewai_tools.BaseTool, implement the _run methodology after which create an occasion of this class.
from crewai_tools import BaseToolclass DatabaseQuery(BaseTool):
title: str = "Database Question"
description: str = "Returns the results of SQL question execution"
def _run(self, sql_query: str) -> str:
# Implementation goes right here
return get_clickhouse_data(sql_query)
database_query_tool = DatabaseQuery()
Now, we will set the duties for the brokers. Once more, offering clear directions and all of the context to LLM is essential.
table_description_task = Process(
description = '''Present the great overview for the information
in desk {desk}, in order that it is simple to grasp the construction
of the information. This activity is essential to place collectively the documentation
for our database''',
expected_output = '''The great overview of {desk} within the md format.
Embrace 2 sections: columns (listing of columns with their sorts)
and examples (the primary 30 rows from desk).''',
instruments = [database_query_tool],
agent = database_specialist_agent
)table_documentation_task = Process(
description = '''Utilizing offered details about the desk,
put collectively the detailed documentation for this desk in order that
individuals can use it in apply''',
expected_output = '''Nicely-written detailed documentation describing
the information scheme for the desk {desk} in markdown format,
that offers the desk overview in 1-2 sentences then then
describes every columm. Construction the columns description
as a markdown desk with column title, sort and outline.''',
instruments = [],
output_file="table_documentation.md",
agent = tech_writer_agent
)
You might need seen that I’ve used {desk} placeholder within the duties’ descriptions. We are going to use desk as an enter variable when executing the crew, and this worth will likely be inserted into all placeholders.
Additionally, I’ve specified the output file for the desk documentation activity to save lots of the ultimate outcome domestically.
We have now all we want. Now, it’s time to create a crew and execute the method, specifying the desk we’re considering. Let’s strive it with the customers desk.
crew = Crew(
brokers = [database_specialist_agent, tech_writer_agent],
duties = [table_description_task, table_documentation_task],
verbose = 2
)outcome = crew.kickoff({'desk': 'ecommerce_db.customers'})
It’s an thrilling second, and I’m actually wanting ahead to seeing the outcome. Don’t fear if execution takes a while. Brokers make a number of LLM calls, so it’s completely regular for it to take a couple of minutes. It took 2.5 minutes on my laptop computer.
We requested LLM to return the documentation in markdown format. We will use the next code to see the formatted end in Jupyter Pocket book.
from IPython.show import Markdown
Markdown(outcome)
At first look, it seems to be nice. We’ve obtained the legitimate markdown file describing the customers’ desk.
However wait, it’s incorrect. Let’s see what information we have now in our desk.
The columns listed within the documentation are fully completely different from what we have now within the database. It’s a case of LLM hallucinations.
We’ve set verbose = 2 to get the detailed logs from CrewAI. Let’s learn via the execution logs to establish the basis reason behind the issue.
First, the database specialist couldn’t question the database as a consequence of issues with quotes.
The specialist didn’t handle to resolve this drawback. Lastly, this chain has been terminated by CrewAI with the next output: Agent stopped as a consequence of iteration restrict or time restrict.
This implies the technical author didn’t obtain any factual details about the information. Nonetheless, the agent continued and produced fully faux outcomes. That’s how we ended up with incorrect documentation.
Fixing the problems
Despite the fact that our first iteration wasn’t profitable, we’ve discovered rather a lot. We have now (no less than) two areas for enchancment:
- Our database instrument is simply too tough for the mannequin, and the agent struggles to make use of it. We will make the instrument extra tolerant by eradicating quotes from the start and finish of the queries. This resolution will not be best since legitimate SQL can finish with a quote, however let’s strive it.
- Our technical author isn’t basing its output on the enter from the database specialist. We have to tweak the immediate to spotlight the significance of offering solely factual info.
So, let’s attempt to repair these issues. First, we’ll repair the instrument — we will leverage strip to remove quotes.
CH_HOST = 'http://localhost:8123' # default handle def get_clickhouse_data(question, host = CH_HOST, connection_timeout = 1500):
r = requests.publish(host, params = {'question': question.strip('"').strip("'")},
timeout = connection_timeout)
if r.status_code == 200:
return r.textual content
else:
return 'Database returned the next error:n' + r.textual content
Then, it’s time to replace the immediate. I’ve included statements emphasizing the significance of sticking to the info in each the agent and activity definitions.
tech_writer_agent = Agent(
position = "Technical author",
purpose = '''Write partaking and factually correct technical documentation
for information sources or instruments''',
backstory = '''
You're an knowledgeable in each expertise and communications, so that you
can simply clarify even subtle ideas.
Your texts are concise and may be simply understood by extensive viewers.
You utilize skilled however somewhat casual type in your communication.
You base your work on the factual info offered by your colleagues.
You stick with the info within the documentation and use ONLY
info offered by the colleagues not including something.''',
allow_delegation = False,
verbose = True
)table_documentation_task = Process(
description = '''Utilizing offered details about the desk,
put collectively the detailed documentation for this desk in order that
individuals can use it in apply''',
expected_output = '''Nicely-written detailed documentation describing
the information scheme for the desk {desk} in markdown format,
that offers the desk overview in 1-2 sentences then then
describes every columm. Construction the columns description
as a markdown desk with column title, sort and outline.
The documentation is predicated ONLY on the knowledge offered
by the database specialist with none additions.''',
instruments = [],
output_file = "table_documentation.md",
agent = tech_writer_agent
)
Let’s execute our crew as soon as once more and see the outcomes.
We’ve achieved a bit higher outcome. Our database specialist was capable of execute queries and think about the information, which is a major win for us. Moreover, we will see all of the related fields within the outcome desk, although there are many different fields as effectively. So, it’s nonetheless not totally right.
I as soon as once more seemed via the CrewAI execution log to determine what went unsuitable. The problem lies in getting the listing of columns. There’s no filter by database, so it returns some unrelated columns that seem within the outcome.
SELECT column_name
FROM information_schema.columns
WHERE table_name = 'customers'
Additionally, after taking a look at a number of makes an attempt, I seen that the database specialist, infrequently, executes choose * from <desk> question. It would trigger some points in manufacturing as it would generate numerous information and ship it to LLM.
Extra specialised instruments
We will present our agent with extra specialised instruments to enhance our resolution. At present, the agent has a instrument to execute any SQL question, which is versatile and highly effective however liable to errors. We will create extra centered instruments, resembling getting desk construction and top-N rows from the desk. Hopefully, it’ll scale back the variety of errors.
class TableStructure(BaseTool):
title: str = "Desk construction"
description: str = "Returns the listing of columns and their sorts"def _run(self, desk: str) -> str:
desk = desk.strip('"').strip("'")
return get_clickhouse_data(
'describe {desk} format TabSeparatedWithNames'
.format(desk = desk)
)
class TableExamples(BaseTool):
title: str = "Desk examples"
description: str = "Returns the primary N rows from the desk"
def _run(self, desk: str, n: int = 30) -> str:
desk = desk.strip('"').strip("'")
return get_clickhouse_data(
'choose * from {desk} restrict {n} format TabSeparatedWithNames'
.format(desk = desk, n = n)
)
table_structure_tool = TableStructure()
table_examples_tool = TableExamples()
Now, we have to specify these instruments within the activity and re-run our script. After the primary try, I obtained the next output from the Technical Author.
Process output: This remaining reply gives an in depth and factual description
of the ecommerce_db.customers desk construction, together with column names, sorts,
and descriptions. The documentation adheres to the offered info
from the database specialist with none additions or modifications.
Extra centered instruments helped the database specialist retrieve the proper desk info. Nonetheless, although the author had all the required info, we didn’t get the anticipated outcome.
As we all know, LLMs are probabilistic, so I gave it one other strive. And hooray, this time, the outcome was fairly good.
It’s not excellent because it nonetheless contains some irrelevant feedback and lacks the general description of the desk. Nonetheless, offering extra specialised instruments has undoubtedly paid off. It additionally helped to stop points when the agent tried to load all the information from the desk.
High quality assurance specialist
We’ve achieved fairly good outcomes, however let’s see if we will enhance them additional. A standard apply in multi-agent setups is high quality assurance, which provides the ultimate overview stage earlier than finalising the outcomes.
Let’s create a brand new agent — a High quality Assurance Specialist, who will likely be accountable for overview.
qa_specialist_agent = Agent(
position = "High quality Assurance specialist",
purpose = """Guarantee the best high quality of the documentation we offer
(that it is right and straightforward to grasp)""",
backstory = '''
You're employed as a High quality Assurance specialist, checking the work
from the technical author and making certain that it is inline
with our highest requirements.
You must test that the technical author gives the total full
solutions and make no assumptions.
Additionally, you might want to guarantee that the documentation addresses
all of the questions and is straightforward to grasp.
''',
allow_delegation = False,
verbose = True
)
Now, it’s time to explain the overview activity. I’ve used the context parameter to specify that this activity requires outputs from each table_description_task and table_documentation_task.
qa_review_task = Process(
description = '''
Evaluation the draft documentation offered by the technical author.
Be certain that the documentation totally solutions all of the questions:
the aim of the desk and its construction within the type of desk.
Ensure that the documentation is according to the knowledge
offered by the database specialist.
Double test that there are not any irrelevant feedback within the remaining model
of documentation.
''',
expected_output = '''
The ultimate model of the documentation in markdown format
that may be printed.
The documentation ought to totally handle all of the questions, be constant
and comply with our skilled however casual tone of voice.
''',
instruments = [],
context = [table_description_task, table_documentation_task],
output_file="checked_table_documentation.md",
agent = qa_specialist_agent
)
Let’s replace our crew and run it.
full_crew = Crew(
brokers=[database_specialist_agent, tech_writer_agent, qa_specialist_agent],
duties=[table_description_task, table_documentation_task, qa_review_task],
verbose = 2,
reminiscence = False # do not work with Llama
)full_result = full_crew.kickoff({'desk': 'ecommerce_db.customers'})
We now have extra structured and detailed documentation due to the addition of the QA stage.
Delegation
With the addition of the QA specialist, it might be attention-grabbing to check the delegation mechanism. The QA specialist agent might need questions or requests that it may delegate to different brokers.
I attempted utilizing the delegation with Llama 3, however it didn’t go effectively. Llama 3 struggled to name the co-worker instrument appropriately. It couldn’t specify the proper co-worker’s title.
We achieved fairly good outcomes with a neighborhood mannequin that may run on any laptop computer, however now it’s time to modify gears and use a far more highly effective mannequin — GPT-4o.
To do it, we simply must replace the next atmosphere variables.
os.environ["OPENAI_MODEL_NAME"] = 'gpt-4o'
os.environ["OPENAI_API_KEY"] = config['OPENAI_API_KEY'] # your OpenAI key
To modify on the delegation, we must always specify allow_delegation = True for the QA specialist agent.
Additionally, we will use useful reminiscence performance for OpenAI fashions, as talked about above. The reminiscence will enable our brokers to share their information with one another throughout execution and leverage long-term reminiscence to get info from earlier executions.
full_crew = Crew(
brokers = [database_specialist_agent, tech_writer_agent, qa_specialist_agent],
duties = [table_description_task, table_documentation_task, qa_review_task],
verbose = 2,
reminiscence = True
)full_result = full_crew.kickoff({'desk': 'ecommerce_db.customers'})
Let’s see the CrewAI execution logs to grasp how delegation works. Listed here are all of the logs for the QA specialist. We will see that it reached out to the database specialist to double-check the knowledge.
Thought: I must overview the offered draft documentation to make sure it
totally solutions all of the questions and is full, right, and straightforward
to grasp. Particularly, I would like to make sure that it explains the aim
of the desk, its construction, and contains related examples with out
any irrelevant feedback.First, I'll evaluate the 2 drafts to establish any discrepancies and
lacking info. Then, I'll confirm the accuracy and completeness of
the knowledge offered.
**Comparability:**
1. **Introduction:**
- First draft: Consists of an introduction and examples.
- Second draft: Provides a extra detailed introduction concerning the goal and significance of the desk.
2. **Columns:**
- Each drafts have the identical column construction and descriptions.
3. **Extra Info:**
- Each drafts embrace comparable further info, however the second draft is extra concise.
4. **Examples:**
- The primary draft contains examples, that are lacking within the second draft.
5. **Remaining Checks:**
- Guarantee there are not any irrelevant feedback.
- Confirm consistency with info offered by the database specialist.
Motion: Ask query to co-worker
Motion Enter:
<-- omitted logs from the database specialist -->
Based mostly on the offered context and examples for the `ecommerce_db.customers` desk,
the column descriptions and examples seem correct. Listed here are the main points:
- `user_id`: This column accommodates a novel identifier for every person.
The offered examples (e.g., 1000001, 1000002) align with this description
as they're distinctive numeric identifiers.
- `nation`: This column specifies the nation of the person. The offered
examples (e.g., United Kingdom, France, Germany, Netherlands) are
all legitimate nation names.
- `is_active`: This column signifies whether or not the person is lively (1) or not (0). The offered examples (e.g., 0, 1) appropriately signify this binary standing.
- `age`: This column exhibits the age of the person. The offered examples (e.g., 70,
87, 88, 25, 48, 78, 65, 31, 66, 73) are all legitimate ages.
Subsequently, the column descriptions and examples offered for the
`ecommerce_db.customers` desk are correct and meet the anticipated standards.
Thought: Based mostly on the statement from the database specialist,
the column descriptions and examples offered for the `ecommerce_db.customers`
desk are correct. Now, I'll consolidate the perfect parts from each drafts,
making certain the documentation is full, clear, and freed from irrelevant feedback.
Motion: Finalize the documentation in markdown format, incorporating
the detailed introduction, examples, and making certain the tone is skilled
however casual.
After I tried the delegation for the primary time, I didn’t allow reminiscence, which led to incorrect outcomes. The info specialist and the technical author initially returned the proper info. Nonetheless, when the QA specialist returned with the follow-up questions, they began to hallucinate. So, it seems to be like delegation works higher when reminiscence is enabled.
Right here’s the ultimate output from GPT-4o. The outcome seems to be fairly good now. We undoubtedly can use LLMs to automate documentation.
So, the primary activity has been solved!
I used the identical script to generate documentation for the ecommerce_db.periods desk as effectively. It will likely be useful for our subsequent activity. So, let’s not waste any time and transfer on.
Our subsequent activity is answering questions based mostly on the documentation because it’s widespread for a lot of information analysts (and different specialists).
We are going to begin easy and can create simply two brokers:
- The documentation assist specialist will likely be answering questions based mostly on the docs,
- The assist QA agent will overview the reply earlier than sharing it with the shopper.
We might want to empower the documentation specialist with a few instruments that may enable them to see all of the information saved within the listing and browse the information. It’s fairly simple since CrewAI has applied such instruments.
from crewai_tools import DirectoryReadTool, FileReadTooldocumentation_directory_tool = DirectoryReadTool(
listing = '~/crewai_project/ecommerce_documentation')
base_file_read_tool = FileReadTool()
Nonetheless, since Llama 3 retains battling quotes when calling instruments, I needed to create a customized instrument on prime of the FileReaderTool to beat this concern.
from crewai_tools import BaseToolclass FileReadToolUPD(BaseTool):
title: str = "Learn a file's content material"
description: str = "A instrument that can be utilized to learn a file's content material."
def _run(self, file_path: str) -> str:
# Implementation goes right here
return base_file_read_tool._run(file_path = file_path.strip('"').strip("'"))
file_read_tool = FileReadToolUPD()
Subsequent, as we did earlier than, we have to create brokers, duties and crew.
data_support_agent = Agent(
position = "Senior Knowledge Assist Agent",
purpose = "Be probably the most useful assist for you colleagues",
backstory = '''You're employed as a assist for data-related questions
within the firm.
Despite the fact that you are a giant knowledgeable in our information warehouse, you double test
all of the info in documentation.
Our documentation is completely up-to-date, so you may totally depend on it
when answering questions (you needn't test the precise information
in database).
Your work is essential for the crew success. Nonetheless, bear in mind
that examples of desk rows do not present all of the potential values.
You must be certain that you present the absolute best assist: answering
all of the questions, making no assumptions and sharing solely the factual information.
Be artistic strive your greatest to unravel the shopper drawback.
''',
allow_delegation = False,
verbose = True
)qa_support_agent = Agent(
position = "Assist High quality Assurance Agent",
purpose = """Guarantee the best high quality of the solutions we offer
to the shoppers""",
backstory = '''You're employed as a High quality Assurance specialist, checking the work
from assist brokers and making certain that it is inline with our highest requirements.
You must test that the agent gives the total full solutions
and make no assumptions.
Additionally, you might want to guarantee that the documentation addresses all
the questions and is straightforward to grasp.
''',
allow_delegation = False,
verbose = True
)
draft_data_answer = Process(
description = '''Crucial buyer {buyer} reached out to you
with the next query:
```
{query}
```
Your activity is to supply the perfect reply to all of the factors within the query
utilizing all obtainable info and never making any assumprions.
If you do not have sufficient info to reply the query, simply say
that you do not know.''',
expected_output = '''The detailed informative reply to the shopper's
query that addresses all the purpose talked about.
Ensure that reply is full and stict to info
(with none further info not based mostly on the factual information)''',
instruments = [documentation_directory_tool, file_read_tool],
agent = data_support_agent
)
answer_review = Process(
description = '''
Evaluation the draft reply offered by the assist agent.
Be certain that the it totally solutions all of the questions talked about
within the preliminary inquiry.
Ensure that the reply is constant and would not embrace any assumptions.
''',
expected_output = '''
The ultimate model of the reply in markdown format that may be shared
with the shopper.
The reply ought to totally handle all of the questions, be constant
and comply with our skilled however casual tone of voice.
We're very chill and pleasant firm, so remember to incorporate
all of the well mannered phrases.
''',
instruments = [],
agent = qa_support_agent
)
qna_crew = Crew(
brokers = [data_support_agent, qa_support_agent],
duties = [draft_data_answer, answer_review],
verbose = 2,
reminiscence = False # do not work with Llama
)
Let’s see the way it works in apply.
outcome = qna_crew.kickoff(
{'buyer': "Max",
'query': """Hey crew, I hope you are doing effectively. I would like to search out
the numbers earlier than our CEO presentation tomorrow, so I'll actually
respect your assist.
I must calculate the variety of periods from our Home windows customers in 2023. I've tried to search out the desk with such information in our information warehouse, however wasn't capable of.
Do you may have any concepts whether or not we retailer the wanted information someplace,
in order that I can question it? """
}
)
We’ve obtained a well mannered, sensible and useful reply in return. That’s actually nice.
**Good day Max,**Thanks for reaching out together with your query! I am comfortable that will help you
discover the variety of periods from Home windows customers in 2023.
After reviewing our documentation, I discovered that we do retailer information
associated to periods and customers in our ecommerce database, particularly in
the `ecommerce_db.periods` desk.
To reply your query, I can give you a step-by-step information
on find out how to question this desk utilizing SQL. First, you should use the `session_id`
column together with the `os` column filtering for "Home windows" and
the `action_date` column filtering for dates in 2023.
Then, you may group the outcomes by `os` utilizing the `GROUP BY` clause
to depend the variety of periods that meet these circumstances.
This is a pattern SQL question that ought to provide the desired output:
```sql
SELECT COUNT(*)
FROM ecommerce_db.periods
WHERE os = 'Home windows'
AND action_date BETWEEN '2023-01-01' AND '2023-12-31'
GROUP BY os;
```
This question will return the whole variety of periods from Home windows
customers in 2023. I hope this helps! You probably have any additional questions or
want extra help, please do not hesitate to ask.
Let’s complicate the duty a bit. Suppose we will get not solely questions on our information but in addition about our instrument (ClickHouse). So, we can have one other agent within the crew — ClickHouse Guru. To offer our CH agent some information, I’ll share a documentation web site with it.
from crewai_tools import ScrapeWebsiteTool, WebsiteSearchTool
ch_documenation_tool = ScrapeWebsiteTool(
'https://clickhouse.com/docs/en/guides/creating-tables')
If you might want to work with a prolonged doc, you would possibly strive utilizing RAG (Retrieval Augmented technology) — WebsiteSearchTool. It is going to calculate embeddings and retailer them domestically in ChromaDB. In our case, we’ll stick with a easy web site scraper instrument.
Now that we have now two subject material consultants, we have to resolve who will likely be engaged on the questions. So, it’s time to make use of a hierarchical course of and add a supervisor to orchestrate all of the duties.
CrewAI gives the supervisor implementation, so we solely must specify the LLM mannequin. I’ve picked the GPT-4o.
from langchain_openai import ChatOpenAI
from crewai import Course ofcomplext_qna_crew = Crew(
brokers = [ch_support_agent, data_support_agent, qa_support_agent],
duties = [draft_ch_answer, draft_data_answer, answer_review],
verbose = 2,
manager_llm = ChatOpenAI(mannequin='gpt-4o', temperature=0),
course of = Course of.hierarchical,
reminiscence = False
)
At this level, I needed to change from Llama 3 to OpenAI fashions once more to run a hierarchical course of because it hasn’t labored for me with Llama (much like this issue).
Now, we will strive our new crew with several types of questions (both associated to our information or ClickHouse database).
ch_result = complext_qna_crew.kickoff(
{'buyer': "Maria",
'query': """Good morning, crew. I am utilizing ClickHouse to calculate
the variety of clients.
May you please remind whether or not there's an choice so as to add totals
in ClickHouse?"""
}
)doc_result = complext_qna_crew.kickoff(
{'buyer': "Max",
'query': """Hey crew, I hope you are doing effectively. I would like to search out
the numbers earlier than our CEO presentation tomorrow, so I'll actually
respect your assist.
I must calculate the variety of periods from our Home windows customers
in 2023. I've tried to search out the desk with such information
in our information warehouse, however wasn't capable of.
Do you may have any concepts whether or not we retailer the wanted information someplace,
in order that I can question it. """
}
)
If we take a look at the ultimate solutions and logs (I’ve omitted them right here since they’re fairly prolonged, but you will discover them and full logs on GitHub), we’ll see that the supervisor was capable of orchestrate appropriately and delegate duties to co-workers with related information to handle the shopper’s query. For the primary (ClickHouse-related) query, we obtained an in depth reply with examples and potential implications of utilizing WITH TOTALS performance. For the data-related query, fashions returned roughly the identical info as we’ve seen above.
So, we’ve constructed a crew that may reply varied kinds of questions based mostly on the documentation, whether or not from a neighborhood file or an internet site. I feel it’s a superb outcome.
You will discover all of the code on GitHub.
On this article, we’ve explored utilizing the CrewAI multi-agent framework to create an answer for writing documentation based mostly on tables and answering associated questions.
Given the intensive performance we’ve utilised, it’s time to summarise the strengths and weaknesses of this framework.
General, I discover CrewAI to be an extremely helpful framework for multi-agent methods:
- It’s simple, and you’ll construct your first prototype shortly.
- Its flexibility permits to unravel fairly subtle enterprise issues.
- It encourages good practices like role-playing.
- It gives many useful instruments out of the field, resembling RAG and an internet site parser.
- The assist of several types of reminiscence enhances the brokers’ collaboration.
- Constructed-in guardrails assist stop brokers from getting caught in repetitive loops.
Nonetheless, there are areas that may very well be improved:
- Whereas the framework is easy and straightforward to make use of, it’s not very customisable. As an illustration, you at present can’t create your individual LLM supervisor to orchestrate the processes.
- Typically, it’s fairly difficult to get the total detailed info from the documentation. For instance, it’s clear that CrewAI applied some guardrails to stop repetitive perform calls, however the documentation doesn’t totally clarify the way it works.
- One other enchancment space is transparency. I like to grasp how frameworks work underneath the hood. For instance, in Langchain, you should use
langchain.debug = Trueto see all of the LLM calls. Nonetheless, I haven’t found out find out how to get the identical stage of element with CrewAI. - The total assist for the native fashions can be an awesome addition, as the present implementation both lacks some options or is tough to get working correctly.
The area and instruments for LLMs are evolving quickly, so I’m hopeful that we’ll see a whole lot of progress within the close to future.
Thank you a large number for studying this text. I hope this text was insightful for you. You probably have any follow-up questions or feedback, please depart them within the feedback part.
This text is impressed by the “Multi AI Agent Systems with CrewAI” quick course from DeepLearning.AI.
