Scribd
Upload
0 views

chapter2

The document discusses developing applications using LangChain, focusing on sequential chains and ReAct agents that utilize large language models (LLMs) to perform tasks. It provides examples of creating prompts, invoking agents, and integrating custom tools for generating outputs like travel itineraries and financial reports. Additionally, it highlights the use of LangGraph for designing agent systems with a unified syntax.

Uploaded by

unkown21
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
0 views

chapter2

The document discusses developing applications using LangChain, focusing on sequential chains and ReAct agents that utilize large language models (LLMs) to perform tasks. It provides examples of creating prompts, invoking agents, and integrating custom tools for generating outputs like travel itineraries and financial reports. Additionally, it highlights the use of LangGraph for designing agent systems with a unified syntax.

Uploaded by

unkown21
Copyright
© © All Rights Reserved
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 28

Sequential chains

D E V E L O P I N G L L M A P P L I C AT I O N S W I T H L A N G C H A I N

Jonathan Bennion
AI Engineer & LangChain Contributor
DEVELOPING LLM APPLICATIONS WITH LANGCHAIN
DEVELOPING LLM APPLICATIONS WITH LANGCHAIN
DEVELOPING LLM APPLICATIONS WITH LANGCHAIN
DEVELOPING LLM APPLICATIONS WITH LANGCHAIN
DEVELOPING LLM APPLICATIONS WITH LANGCHAIN
DEVELOPING LLM APPLICATIONS WITH LANGCHAIN
Sequential chains
Output → input

destination_prompt = PromptTemplate(
input_variables=["destination"],
template="I am planning a trip to {destination}. Can you suggest some activities to do there?"
)
activities_prompt = PromptTemplate(
input_variables=["activities"],
template="I only have one day, so can you create an itinerary from your top three activities: {activities}."
)

llm = ChatOpenAI(model="gpt-4o-mini", api_key=openai_api_key)

seq_chain = ({"activities": destination_prompt | llm | StrOutputParser()}


| activities_prompt
| llm
| StrOutputParser())

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


print(seq_chain.invoke({"destination": "Rome"}))

- Morning:
1. Start your day early with a visit to the Colosseum. Take a guided tour to learn about its history and significance.
2. After exploring the Colosseum, head to the Roman Forum and Palatine Hill to see more of ancient Rome's ruins.

- Lunch:
3. Enjoy a delicious Italian lunch at a local restaurant near the historic center.

- Afternoon:
4. Visit the Vatican City and explore St. Peter's Basilica, the Vatican Museums, and the Sistine Chapel.
5. Take some time to wander through the charming streets of Rome, stopping at landmarks like the Pantheon, Trevi
Fountain, and Piazza Navona.

- Evening:
6. Relax in one of Rome's beautiful parks, such as Villa Borghese or the Orange Garden, for a peaceful escape from the
bustling city.
7. End your day with a leisurely dinner at a local restaurant, indulging in more Italian cuisine and maybe some gelato.

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Let's practice!
D E V E L O P I N G L L M A P P L I C AT I O N S W I T H L A N G C H A I N
Introduction to
LangChain agents
D E V E L O P I N G L L M A P P L I C AT I O N S W I T H L A N G C H A I N

Jonathan Bennion
AI Engineer & LangChain Contributor
What are agents?

Agents: use LLMs to take actions

Tools: functions called by the agent

Now → ReAct Agent

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


ReAct agents
Reason + Act

What is the weather like in Kingston,


Jamaica?

Thought: I should call Weather() to find the


weather in Kingston, Jamaica.

Act: Weather("Kingston, Jamaica")

Observe: The weather is mostly sunny with


temperatures of 82°F.

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


LangGraph

Branch of LangChain centered around


designing agent systems

Unified, tool-agnostic syntax

pip install langgraph==0.066

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


ReAct agent
from langgraph.prebuilt import create_react_agent
from langchain.agents import load_tools

llm = ChatOpenAI(model="gpt-4o-mini", api_key=openai_api_key)


tools = load_tools(["llm-math"], llm=llm)
agent = create_react_agent(llm, tools)

messages = agent.invoke({"messages": [("human", "What is the square root of 101?")]})


print(messages)

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


ReAct agent
{'messages': [
HumanMessage(content='What is the square root of 101?', ...),
AIMessage(content='', ..., tool_calls=[{'name': 'Calculator', 'args': {'__arg1': 'sqrt(101)'}, ...),
ToolMessage(content='Answer: 10.04987562112089', ...),
AIMessage(content='The square root of 101 is approximately 10.05.', ...)
]}

print(messages['messages'][-1].content)

The square root of 101 is approximately 10.05.

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Let's practice!
D E V E L O P I N G L L M A P P L I C AT I O N S W I T H L A N G C H A I N
Custom tools for
agents
D E V E L O P I N G L L M A P P L I C AT I O N S W I T H L A N G C H A I N

Jonathan Bennion
AI Engineer & LangChain Contributor
Tool formats
from langchain.agents import load_tools

tools = load_tools(["llm-math"], llm=llm)


print(tools[0].name)

Calculator

print(tools[0].description)

Useful for when you need to answer questions about math.

Used by LLM/agent as context to determine when to call it

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Tool formats
print(tools[0].return_direct)

False

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Defining a custom function
def financial_report(company_name: str, revenue: int, expenses: int) -> str:
"""Generate a financial report for a company that calculates net income."""
net_income = revenue - expenses

report = f"Financial Report for {company_name}:\n"


report += f"Revenue: ${revenue}\n"
report += f"Expenses: ${expenses}\n"
report += f"Net Income: ${net_income}\n"
return report

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Calling the function
print(financial_report(company_name="LemonadeStand", revenue=100, expenses=50))

Financial Report for LemonadeStand:


Revenue: $100
Expenses: $50
Net Income: $50

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


From functions to tools
from langchain_core.tools import tool

@tool
def financial_report(company_name: str, revenue: int, expenses: int) -> str:
"""Generate a financial report for a company that calculates net income."""
net_income = revenue - expenses

report = f"Financial Report for {company_name}:\n"


report += f"Revenue: ${revenue}\n"
report += f"Expenses: ${expenses}\n"
report += f"Net Income: ${net_income}\n"
return report

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Examining our new tool
print(financial_report.name)
print(financial_report.description)
print(financial_report.return_direct)
print(financial_report.args)

financial_report
Generate a financial report for a company that calculates net income.
False
{'company_name': {'title': 'Company Name', 'type': 'string'},
'revenue': {'title': 'Revenue', 'type': 'integer'},
'expenses': {'title': 'Expenses', 'type': 'integer'}}

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Integrating the custom tool
from langgraph.prebuilt import create_react_agent

llm = ChatOpenAI(model="gpt-4o-mini", api_key=openai_api_key, temperature=0)


agent = create_react_agent(llm, [financial_report])

messages = agent.invoke({"messages": [("human", "TechStack generated made $10 million


with $8 million of costs. Generate a financial report.")]})
print(messages)

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Integrating the custom tool
{'messages': [
HumanMessage(content='TechStack generated made $10 million dollars with $8 million of...', ...),
AIMessage(content='', ..., tool_calls=[{'name': 'financial_report',
'args': {'company_name': 'TechStack',
'revenue': 10000000, 'expenses': 8000000}, ...),
ToolMessage(content='Financial Report for TechStack:\nRevenue: $10000000\nExpenses...', ...),
AIMessage(content='Here is the financial report for TechStack...', ...)
]}

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Tool outputs
print(messages['messages'][-1].content)

Here is the financial report for TechStack:


- Revenue: $10,000,000
- Expenses: $8,000,000
- Net Income: $2,000,000

Financial Report for TechStack:


Revenue: $10000000
Expenses: $8000000
Net Income: $2000000

DEVELOPING LLM APPLICATIONS WITH LANGCHAIN


Let's practice!
D E V E L O P I N G L L M A P P L I C AT I O N S W I T H L A N G C H A I N

You might also like