Use Cases

End-to-end patterns built from verified Buddy AI APIs. Each combines a few core pieces — model, tools, knowledge, memory, or a team — into something useful.

Research assistant

An agent that searches the web, runs calculations, and writes up findings.

from buddy import Agent
from buddy.models.openai import OpenAIChat
from buddy.tools.tavily import TavilyTools
from buddy.tools.python import PythonTools

assistant = Agent(
    name="research_assistant",
    model=OpenAIChat(id="gpt-4o"),
    tools=[TavilyTools(), PythonTools()],
    instructions=[
        "Search the web for current, credible sources.",
        "Verify any numbers by running Python.",
        "Summarize findings with a short bulleted list and cite sources.",
    ],
    show_tool_calls=True,
    markdown=True,
)

assistant.print_response("Summarize the state of solar panel efficiency in 2026.")

Keys

Needs OPENAI_API_KEY and TAVILY_API_KEY (pip install tavily-python).

Customer-support bot with memory

A support agent that remembers each user across turns using persistent memory and session storage.

from buddy import Agent
from buddy.models.openai import OpenAIChat
from buddy.memory.v2.memory import Memory
from buddy.memory.v2.db.sqlite import SqliteMemoryDb
from buddy.storage.sqlite import SqliteStorage

support = Agent(
    name="support_bot",
    model=OpenAIChat(id="gpt-4o-mini"),
    memory=Memory(
        model=OpenAIChat(id="gpt-4o-mini"),
        db=SqliteMemoryDb(table_name="support_memories", db_file="tmp/support.db"),
    ),
    enable_user_memories=True,
    storage=SqliteStorage(table_name="support_sessions", db_file="tmp/support.db"),
    add_history_to_messages=True,
    num_history_runs=5,
    instructions=[
        "You are a friendly customer-support agent.",
        "Use what you remember about the user to personalize replies.",
        "If you cannot resolve an issue, summarize it for a human agent.",
    ],
    markdown=True,
)

support.print_response("Hi, my order #4471 hasn't arrived.", user_id="cust_88", session_id="ticket_4471")
support.print_response("Any update on that?", user_id="cust_88", session_id="ticket_4471")

Why it works: Memory + enable_user_memories captures durable facts about the user; storage persists the session so a conversation can be resumed by its session_id; add_history_to_messages keeps recent turns in context.

Document Q&A over PDFs

Index a folder of PDFs and answer questions grounded in their content.

from buddy import Agent
from buddy.models.openai import OpenAIChat
from buddy.knowledge.pdf import PDFKnowledgeBase
from buddy.vectordb.chroma import ChromaDb

knowledge = PDFKnowledgeBase(
    path="docs/policies/",  # a PDF file or a directory of PDFs
    vector_db=ChromaDb(collection="policies", persistent_client=True),
)
knowledge.load()  # run once; re-run with recreate=True to rebuild

qa = Agent(
    model=OpenAIChat(id="gpt-4o-mini"),
    knowledge=knowledge,
    search_knowledge=True,
    instructions=["Answer only from the documents.", "Quote the relevant passage."],
    markdown=True,
)

qa.print_response("What is the refund policy for damaged items?")

Filtered retrieval

If you attached metadata to documents, restrict retrieval with qa.run(question, knowledge_filters={"category": "returns"}).

Data analysis agent

An agent that writes and runs Python to analyze data and report results.

from buddy import Agent
from buddy.models.openai import OpenAIChat
from buddy.tools.python import PythonTools
from buddy.tools.calculator import CalculatorTools

analyst = Agent(
    name="data_analyst",
    model=OpenAIChat(id="gpt-4o"),
    tools=[PythonTools(), CalculatorTools()],
    instructions=[
        "Write Python to load and analyze the data.",
        "Show the code and the computed result.",
        "Explain findings in plain language.",
    ],
    show_tool_calls=True,
    markdown=True,
)

analyst.print_response(
    "Given monthly revenue [120, 135, 150, 128, 162], compute the mean, "
    "the month-over-month growth, and the overall trend."
)

What to expect: the agent emits Python via PythonTools, runs it, and returns the numeric results alongside an explanation.

Structured extraction pipeline

Combine tools with response_model to return validated, typed data.

from pydantic import BaseModel
from buddy import Agent
from buddy.models.openai import OpenAIChat

class Contact(BaseModel):
    name: str
    email: str
    company: str | None = None

extractor = Agent(model=OpenAIChat(id="gpt-4o"), response_model=Contact)

result = extractor.run("Reach out to Dana Lee at dana@acme.io from Acme Corp.")
contact = result.content  # a validated Contact instance
print(contact.name, contact.email, contact.company)