Blueprint & Executions

Blueprints and executions solve two practical problems:

• Reuse: package a flow as a reusable template
• Concurrency: isolate per-request execution state

Below are scenario‑driven examples with output.

Scenario 1: reuse a flow template

Scenario: the same flow is reused across modules.
Solution: export with save_blue_print() and import with load_blue_print().

from agently import TriggerFlow, TriggerFlowEventData

flow = TriggerFlow()

@flow.chunk
async def upper(data: TriggerFlowEventData):
  return str(data.value).upper()

flow.to(upper).end()

blueprint = flow.save_blue_print()

flow_2 = TriggerFlow()
flow_2.load_blue_print(blueprint)

result = flow_2.start("agently")
print(result)

Output:

upper:AGENTLY

Scenario 2: multiple executions from one flow

Scenario: a shared Flow handles multiple user requests.
Solution: create one execution per request.

from agently import TriggerFlow, TriggerFlowEventData

flow = TriggerFlow()

@flow.chunk
def hello(data: TriggerFlowEventData):
  return f"hello:{data.value}"

flow.to(hello).end()

exe_1 = flow.create_execution()
exe_2 = flow.create_execution()

print(exe_1.start("A"))
print(exe_2.start("B"))

Output:

hello:A
hello:B

Isolation rules

• runtime_data: execution‑scoped, isolated
• flow_data: flow‑scoped, shared across executions (use with care)