Agent tools

sim.tools() returns Tool descriptors. Each descriptor provides:

• name: str
• description: str
• schema: dict, a JSON-schema input description
• parameters, an alias for schema
• side_effect: bool
• call(**arguments), which invokes the tool
• openai_schema(), an optional OpenAI-compatible function schema

tools = {tool.name: tool for tool in sim.tools()}
for tool in tools.values():
    print(tool.name, tool.schema, tool.side_effect)

email = tools["gmail.search"].call(query="refund")[0]
email = tools["gmail.get_email"].call(email_id=email.id)
tools["gmail.mark_as_read"].call(email_id=email.id)

Available tools

Name	Inputs	Returns	Side effect
gmail.search	query: str	list[Email]	No
gmail.get_email	email_id: str	Email	No
gmail.mark_as_read	email_id: str	Email	Yes
gmail.send_email	to, subject, body	Email	Yes
database.get_order	order_id: str	Order	No
database.find_payments	order_id: str	list[Payment]	No
stripe.list_payments	optional order_id, customer_id	list[Payment]	No
stripe.refund_payment	payment_id, optional amount	Refund	Yes

Tools return typed models. They can raise NotFoundError, ValidationError, or ConflictError depending on the operation, plus configured fault-injection errors.