Lab 7 — The Same Agent, with Frameworks¶

⏱️ ~35 min

Part A: rebuild Lab 5's agent with LangChain. Part B: a 2-agent CrewAI team. Uses OpenAI.

Requires an OpenAI API key — these labs use OpenAI embeddings / function-calling specifically. (Labs 1–3 work with OpenAI or Anthropic.)

In [ ]:

# Tip: pin these to exact versions in real projects — these frameworks ship breaking changes often.
!pip install -q langchain langchain-openai crewai
import os, getpass
if not os.environ.get("OPENAI_API_KEY"):
    os.environ["OPENAI_API_KEY"] = getpass.getpass("OpenAI API key: ")

# Tip: pin these to exact versions in real projects — these frameworks ship breaking changes often. !pip install -q langchain langchain-openai crewai import os, getpass if not os.environ.get("OPENAI_API_KEY"): os.environ["OPENAI_API_KEY"] = getpass.getpass("OpenAI API key: ")

Part A — LangChain agent with two tools¶

In [ ]:

import ast, operator
from langchain_core.tools import tool
from langchain_openai import ChatOpenAI
from langchain.agents import create_tool_calling_agent, AgentExecutor
from langchain_core.prompts import ChatPromptTemplate

_OPS = {ast.Add: operator.add, ast.Sub: operator.sub, ast.Mult: operator.mul,
        ast.Div: operator.truediv, ast.Pow: operator.pow, ast.Mod: operator.mod,
        ast.USub: operator.neg}
def _safe_arith(node):
    if isinstance(node, ast.Constant) and isinstance(node.value, (int, float)):
        return node.value
    if isinstance(node, ast.BinOp) and type(node.op) in _OPS:
        return _OPS[type(node.op)](_safe_arith(node.left), _safe_arith(node.right))
    if isinstance(node, ast.UnaryOp) and type(node.op) in _OPS:
        return _OPS[type(node.op)](_safe_arith(node.operand))
    raise ValueError("unsupported expression")

@tool
def web_search(query: str) -> str:
    """Search the web for a fact."""
    facts = {"eiffel tower height": "The Eiffel Tower is 330 meters tall."}
    for k, v in facts.items():
        if k in query.lower():
            return v
    return "No results found."

@tool
def calculator(expression: str) -> str:
    """Safely evaluate a simple arithmetic expression.
    NOTE: never call Python eval() on model-generated text — that's how you get
    code injection. We parse a tiny arithmetic grammar with ast instead."""
    try:
        return str(_safe_arith(ast.parse(expression, mode="eval").body))
    except Exception as e:
        return f"Error: {e}"

prompt = ChatPromptTemplate.from_messages([
    ("system", "You are a helpful agent. Use tools when needed."),
    ("human", "{input}"),
    ("placeholder", "{agent_scratchpad}"),
])
llm = ChatOpenAI(model="gpt-4o-mini", temperature=0)
agent = create_tool_calling_agent(llm, [web_search, calculator], prompt)
executor = AgentExecutor(agent=agent, tools=[web_search, calculator], verbose=True)

import ast, operator from langchain_core.tools import tool from langchain_openai import ChatOpenAI from langchain.agents import create_tool_calling_agent, AgentExecutor from langchain_core.prompts import ChatPromptTemplate _OPS = {ast.Add: operator.add, ast.Sub: operator.sub, ast.Mult: operator.mul, ast.Div: operator.truediv, ast.Pow: operator.pow, ast.Mod: operator.mod, ast.USub: operator.neg} def _safe_arith(node): if isinstance(node, ast.Constant) and isinstance(node.value, (int, float)): return node.value if isinstance(node, ast.BinOp) and type(node.op) in _OPS: return _OPS[type(node.op)](_safe_arith(node.left), _safe_arith(node.right)) if isinstance(node, ast.UnaryOp) and type(node.op) in _OPS: return _OPS[type(node.op)](_safe_arith(node.operand)) raise ValueError("unsupported expression") @tool def web_search(query: str) -> str: """Search the web for a fact.""" facts = {"eiffel tower height": "The Eiffel Tower is 330 meters tall."} for k, v in facts.items(): if k in query.lower(): return v return "No results found." @tool def calculator(expression: str) -> str: """Safely evaluate a simple arithmetic expression. NOTE: never call Python eval() on model-generated text — that's how you get code injection. We parse a tiny arithmetic grammar with ast instead.""" try: return str(_safe_arith(ast.parse(expression, mode="eval").body)) except Exception as e: return f"Error: {e}" prompt = ChatPromptTemplate.from_messages([ ("system", "You are a helpful agent. Use tools when needed."), ("human", "{input}"), ("placeholder", "{agent_scratchpad}"), ]) llm = ChatOpenAI(model="gpt-4o-mini", temperature=0) agent = create_tool_calling_agent(llm, [web_search, calculator], prompt) executor = AgentExecutor(agent=agent, tools=[web_search, calculator], verbose=True)

In [ ]:

executor.invoke({"input": "How tall is the Eiffel Tower, and what is that doubled?"})

executor.invoke({"input": "How tall is the Eiffel Tower, and what is that doubled?"})

Compare to Lab 5: no manual message list, no tool-dispatch loop — the executor runs it.

Part B — A two-agent team with CrewAI¶

A researcher finds the fact; a math analyst doubles it. Shows multi-agent orchestration.

In [ ]:

from crewai import Agent, Task, Crew

researcher = Agent(
    role="Researcher", goal="Find the height of the Eiffel Tower",
    backstory="You find accurate facts.", verbose=True, allow_delegation=False)
analyst = Agent(
    role="Math Analyst", goal="Double any number you are given",
    backstory="You do precise arithmetic.", verbose=True, allow_delegation=False)

t1 = Task(description="State the Eiffel Tower's height in meters.",
          expected_output="A single number in meters.", agent=researcher)
t2 = Task(description="Double the height the researcher found.",
          expected_output="The doubled number.", agent=analyst, context=[t1])

crew = Crew(agents=[researcher, analyst], tasks=[t1, t2], verbose=True)
print(crew.kickoff())

from crewai import Agent, Task, Crew researcher = Agent( role="Researcher", goal="Find the height of the Eiffel Tower", backstory="You find accurate facts.", verbose=True, allow_delegation=False) analyst = Agent( role="Math Analyst", goal="Double any number you are given", backstory="You do precise arithmetic.", verbose=True, allow_delegation=False) t1 = Task(description="State the Eiffel Tower's height in meters.", expected_output="A single number in meters.", agent=researcher) t2 = Task(description="Double the height the researcher found.", expected_output="The doubled number.", agent=analyst, context=[t1]) crew = Crew(agents=[researcher, analyst], tasks=[t1, t2], verbose=True) print(crew.kickoff())

Takeaway: LangChain abstracts the single-agent loop; CrewAI coordinates multiple specialist agents. Frameworks trade some control for far less boilerplate.

✅ Did it work?¶

• Part A: the verbose AgentExecutor log showed it calling web_search then calculator, ending at 660 meters.
• Part B: the CrewAI run printed the researcher's height (330 m) handed to the analyst, whose final output is the doubled value.