Agent

from cachy import enable_cachy
enable_cachy()

m = "anthropic/claude-haiku-4-5"

tool_ctx is injected into tools automatically, and shouldn’t be included in the schema passed to the llm

source

mk_schema


def mk_schema(
    fn, strip:tuple=('tool_ctx',)
):

Call self as a function.

source

Agent


def Agent(
    model:str, # model str in litellm format
    name:str='', # name of the agent
    desc:str='', # what the agent does (used by parent agents to decide when to transfer)
    sp:NoneType=None, # the system prompt
    tools:NoneType=None, # a list of functions to provide to the agent
    subagents:NoneType=None, # a list subagents that this agent can transfer to
    narrative_cast:bool=True, # to narrative cast or not
    include_hist:bool=True, # if False: will remove all history that doesn't belong to this agent on call
    before_cb:NoneType=None, # a callback func(state:dict, msgs:list). Any non-None return will skip the llm
    after_cb:NoneType=None, # a callback func(state:dict, response) that can edit and modify the llm response
    output_key:NoneType=None, # saves model response to state with a keyname defined by output_key
):

Initialize self. See help(type(self)) for accurate signature.

The most simple call function. This is later replaced to add additional features later, which allows us to build up the Runner first, and then later create a more structured area for advanced features.

@patch
def __call__(self:Agent, msgs:list, state=None, **kwargs):
    if self.sp: msgs = [{"content": self.sp, "role": "system"}, *msgs]
    return completion(model=self.model, messages=msgs, tools=self.tool_schemas, **kwargs)

Runner

source

call_llm_tool


def call_llm_tool(
    tc_func:dict, tool_map:dict, state:dict
):

Call self as a function.

source

mk_transfer_doc


def mk_transfer_doc(
    agent, parent
):

Call self as a function.

source

with_model_nm


def with_model_nm(
    llm_output, nm:str
):

Call self as a function.

source

Runner


def Runner(
    agent:Agent
):

Initialize self. See help(type(self)) for accurate signature.

Full agent example (1)

def mult(a:int, b:int):
    "Multiplies two numbers together"
    return a * b

def save(result:str, tool_ctx):
    "Saves some result passed into into storage"
    tool_ctx["final_result"] = result

c1 = Agent(m, name="saver", desc="Saves results into storage.", sp="You recieve results from other agents and save them with your tool.", tools=[save])
a1 = Agent(m, name="math_agent", desc="Is good at doing maths.", sp="Always do maths using tools and always pass to saver to save results.", tools=[mult], subagents=[c1])
a2 = Agent(m, name="root", desc="General agent", sp="A general coordinator that delegates to sub agents as needed", subagents=[a1])

r = Runner(a2)
r("What's 2592321 x 235183")

🔧 transfer_to_agent({“agent_name”: “math_agent”})

id: chatcmpl-88c76514-666c-43ea-977e-c22426dc3dff
model: claude-haiku-4-5-20251001
finish_reason: tool_calls
usage: Usage(completion_tokens=74, prompt_tokens=614, total_tokens=688, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=74, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

🔧 mult({“a”: 2592321, “b”: 235183})

id: chatcmpl-5449c005-c701-4223-9c0e-67f5a4e610c8
model: claude-haiku-4-5-20251001
finish_reason: tool_calls
usage: Usage(completion_tokens=80, prompt_tokens=790, total_tokens=870, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=80, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

🔧 transfer_to_agent({“agent_name”: “saver”})

id: chatcmpl-ed04a5b5-4173-4970-ad58-60da19e14f4c
model: claude-haiku-4-5-20251001
finish_reason: tool_calls
usage: Usage(completion_tokens=66, prompt_tokens=886, total_tokens=952, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=66, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

2592321 × 235183 = 609,669,829,743

Would you like me to save this result?

id: chatcmpl-dc914dde-6f8b-4855-834e-96c1de6a2c47
model: claude-haiku-4-5-20251001
finish_reason: stop
usage: Usage(completion_tokens=47, prompt_tokens=941, total_tokens=988, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=47, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

Choices(finish_reason='stop', index=0, message=Message(content='I apologize for the confusion above. Let me calculate this directly:\n\n2592321 × 235183 = **609,669,829,743**\n\nWould you like me to save this result?', role='assistant', tool_calls=None, function_call=None, provider_specific_fields={'citations': None, 'thinking_blocks': None}))

r.state

{}

Note: although the agent transfering and tool calls are working, the LLM gets confused when seeing messages from previous agents, thinking that it was the one that made those responses. - This is fixed with narrative casting (below)

Advanced features

test_msgs = [
    {"role": "user", "content": "What can you help me with?"},
    {"role": "assistant", "agent": "main", "content": "Ask me any question you want."},
    {"role": "user", "content": "What's the weather?"},
    {"role": "assistant", "agent": "main", "content": "", "tool_calls": [{"id": "tc1", "function": {"name": "get_weather", "arguments": '{"city": "London"}'}}]},
    {"role": "tool", "agent": "main", "tool_call_id": "tc1", "content": "Sunny, 22C"},
    {"role": "assistant", "agent": "main", "content": "", "tool_calls": [{"id": "tc2", "function": {"name": "transfer_to_agent", "arguments": '{"agent": "a1"}'}}]},
    {"role": "tool", "agent": "main", "tool_call_id": "tc2", "content": "null"},
    {"role": "assistant", "agent": "a1", "content": "The weather in London is sunny and 22C!"},
]

Preprocessing

Narrative casting

Narrative casting simply means to transform messages from prior agents to be: - Summarized - Tagged as coming from different agents

source

mk_narrative_cast


def mk_narrative_cast(
    msg_list:list, active_agent:str
):

Call self as a function.

source

msg_fragment


def msg_fragment(
    d:dict, results:dict
):

Call self as a function.

History filtering

For a long set of agent calls, it can be worthwhile for some subagent to ignore most of the history in order to preserve context.

One pattern for agent communication is to use state instead of history: - Agent 1 sets state that agent 2 needs - Agent 2 reads that state (with automatic state insertion below) and sets state for agent 3

source

filter_out_history


def filter_out_history(
    msgs:list, active_agent:str
):

Call self as a function.

filter_out_history(test_msgs[:1], "a1")

[{'role': 'user', 'content': 'What can you help me with?'}]

filter_out_history(test_msgs[:2], "a1")

[{'role': 'assistant',
  'agent': 'main',
  'content': 'Ask me any question you want.'}]

filter_out_history(test_msgs[:3], "a1")

[{'role': 'user', 'content': "What's the weather?"}]

filter_out_history(test_msgs[:6], "a1")

[{'role': 'assistant',
  'agent': 'main',
  'content': '',
  'tool_calls': [{'id': 'tc2',
    'function': {'name': 'transfer_to_agent',
     'arguments': '{"agent": "a1"}'}}]}]

Preproc

source

Agent.preproc_msg


def preproc_msg(
    msgs:list | str, state:NoneType=None
):

Call self as a function.

# Simple str input
Agent(m).preproc_msg("test")

[{'content': 'test', 'role': 'user'}]

# System prompt (with state insertion)
Agent(m, name="a1", sp="You help {user} to do {task}").preproc_msg("test", state={"user": "PersonA", "task": "maths", "extra": 5})

[{'content': 'You help PersonA to do maths', 'role': 'system'},
 {'content': 'test', 'role': 'user'}]

# Narrative casting (default)
Agent(m, name="a1").preproc_msg(test_msgs)

[{'role': 'user', 'content': 'What can you help me with?'},
 {'role': 'user', 'content': "[Agent 'main'] Ask me any question you want."},
 {'role': 'user', 'content': "What's the weather?"},
 {'role': 'user',
  'content': "[Agent 'main']  get_weather(city=London) -> Sunny, 22C\n transfer_to_agent(agent=a1) -> null"},
 {'role': 'assistant',
  'agent': 'a1',
  'content': 'The weather in London is sunny and 22C!'}]

# Don't include full history for subagents
Agent(m, name="a1", include_hist=False).preproc_msg(test_msgs)

[{'role': 'user',
  'content': "[Agent 'main']  transfer_to_agent(agent=a1) -> null"},
 {'role': 'assistant',
  'agent': 'a1',
  'content': 'The weather in London is sunny and 22C!'}]

Callbacks and outputkey

source

mk_cb_msg


def mk_cb_msg(
    msg:str | functools.partial | litellm.types.utils.ModelResponse
):

Call self as a function.

source

tool_response


def tool_response(
    name:str, arguments:dict
):

Call self as a function.

source

assistant_response


def assistant_response(
    content:str
):

Call self as a function.

source

Agent.run_callback


def run_callback(
    fn, msgs:list, state:dict
):

Call self as a function.

source

Agent.maybe_save_to_output_key


def maybe_save_to_output_key(
    state, response
):

Call self as a function.

source

Agent.call


def __call__(
    msgs:list | str, state:NoneType=None, kwargs:VAR_KEYWORD
):

Call self as a function.

# Before callback returning str
def skip_with_str(state, msgs): return "Skipped from before_cbs"
Agent(m, before_cb=skip_with_str)("Test")

Skipped from before_cbs

id: chatcmpl-0967fb02-ac50-4a56-8c44-0082a7a53e53
model: None
finish_reason: stop

# Before callback returning partial function
def skip_with_partial(state, msgs): return partial(mult, a=3, b=5)
Agent(m, before_cb=skip_with_partial)("Test")

🔧 mult({“a”: 3, “b”: 5})

id: chatcmpl-ff8f5441-9d3f-4643-b4b4-2aa1d5086ef0
model: None
finish_reason: tool_calls

# Before callback modifying msgs
def modify_msgs(state, msgs): msgs[-1]["content"] = "What is 5 - 1?"
Agent(m, before_cb=modify_msgs)([{"role": "user", "content": "Test"}])

5 - 1 = 4

id: chatcmpl-3e0be55c-2aa6-4acc-9d2d-1727b982853c
model: claude-haiku-4-5-20251001
finish_reason: stop
usage: Usage(completion_tokens=13, prompt_tokens=16, total_tokens=29, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=13, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

# After callback replacing result
def replace_res(state, response): return assistant_response("I have replaced the response!")
Agent(m, after_cb=replace_res)("Test")

I have replaced the response!

id: chatcmpl-786e97c5-273a-4e47-8172-10d7687cb60f
model: None
finish_reason: stop

# Use before callback to populate state

def get_initial_state(state, msgs):
    state["val"] = 5
    state["name"] = "Smoke"

a = Agent(m, name="poemer", sp="Make a tiny poem about val={val} and name={name}", before_cb=get_initial_state)
Runner(a)("Start")

Smoke drifts through air so free, With val of five, light as can be, A wispy name, a gentle way, Val=5, Smoke holds sway.

id: chatcmpl-66e8d992-7e6d-461c-8e4b-5bb9641b312f
model: claude-haiku-4-5-20251001
finish_reason: stop
usage: Usage(completion_tokens=51, prompt_tokens=22, total_tokens=73, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=51, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

Choices(finish_reason='stop', index=0, message=Message(content="# Smoke's Number\n\nSmoke drifts through air so free,\nWith val of five, light as can be,\nA wispy name, a gentle way,\nVal=5, Smoke holds sway.", role='assistant', tool_calls=None, function_call=None, provider_specific_fields={'citations': None, 'thinking_blocks': None}))

Full agent example (2)

r2 = Runner(a2)
r2("What's 359231 x 235981")

🔧 transfer_to_agent({“agent_name”: “math_agent”})

id: chatcmpl-3bd94f0d-02c2-48b9-9bfc-143671d1818c
model: claude-haiku-4-5-20251001
finish_reason: tool_calls
usage: Usage(completion_tokens=60, prompt_tokens=613, total_tokens=673, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=60, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

🔧 mult({“a”: 359231, “b”: 235981})

id: chatcmpl-e3eebccc-ce1f-4571-87f1-65cf84cffc97
model: claude-haiku-4-5-20251001
finish_reason: tool_calls
usage: Usage(completion_tokens=86, prompt_tokens=725, total_tokens=811, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=86, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

Now let me save this result:

🔧 transfer_to_agent({“agent_name”: “saver”})

id: chatcmpl-f2b764c5-cf9c-479d-8c25-80bd492358d9
model: claude-haiku-4-5-20251001
finish_reason: tool_calls
usage: Usage(completion_tokens=81, prompt_tokens=827, total_tokens=908, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=81, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

🔧 save({“result”: “359231 0d7 235981 = 84771690611”})

id: chatcmpl-850f306c-27a0-4747-adc2-87acbd249524
model: claude-haiku-4-5-20251001
finish_reason: tool_calls
usage: Usage(completion_tokens=65, prompt_tokens=780, total_tokens=845, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=65, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

id: chatcmpl-897fe0b2-203e-40c2-bfe6-dc1ba14194a9
model: claude-haiku-4-5-20251001
finish_reason: stop
usage: Usage(completion_tokens=37, prompt_tokens=858, total_tokens=895, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=37, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

Choices(finish_reason='stop', index=0, message=Message(content="Perfect! I've saved the result. The multiplication of **359231 × 235981 = 84,771,690,611** has been stored.", role='assistant', tool_calls=None, function_call=None, provider_specific_fields={'citations': None, 'thinking_blocks': None}))

r2.state

{'final_result': '359231 × 235981 = 84771690611'}

Sequential

source

Sequential


def Sequential(
    agent_list:list
):

Initialize self. See help(type(self)) for accurate signature.

a1s = Agent(m, name="mathjoke", sp="Your job is to say a maths joke")
a2s = Agent(m, name="sciencsejoke", sp="Your job is to say a science joke")
a3s = Agent(m, name="complainer", sp="You will see context from other agents that are saying jokes. Your job is to complain about them.")

s = Sequential([a1s, a2s, a3s])
s(start_msg="start now")

Because it knew it wasn’t greater than or less than anyone else! 😄

id: chatcmpl-37010874-f3ef-49f9-b1fb-548c94d7ad51
model: claude-haiku-4-5-20251001
finish_reason: stop
usage: Usage(completion_tokens=29, prompt_tokens=18, total_tokens=47, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=29, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

Because it was traveling light! 💡

id: chatcmpl-7c2dab86-0c27-491c-94e2-ff781628d451
model: claude-haiku-4-5-20251001
finish_reason: stop
usage: Usage(completion_tokens=27, prompt_tokens=51, total_tokens=78, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=27, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

That math joke? “The equal sign wasn’t greater than or less than anyone else”? That’s not clever, that’s just… painful. It’s forcing a pun out of basic mathematical symbols. We get it, equal sign = equality metaphor. Where’s the actual humor? This feels like something a textbook would say to try to make algebra “fun” for bored middle schoolers.

And don’t even get me started on the photon joke. “Traveling light” because photons ARE light? That’s the laziest setup-and-punchline I’ve heard all week. There’s no wit, no timing, no surprise twist. It’s just taking a literal property of photons and calling it a joke. A hotel joke that doesn’t even make sense logically—why would a photon need to check into a hotel in the first place?

Both of these feel like they were generated by someone who just learned what puns are and decided that any word-double-meaning automatically equals comedy. Hard pass on both. 😤

id: chatcmpl-d60b8526-0e9c-4bdf-9315-4a19d630ad0f
model: claude-haiku-4-5-20251001
finish_reason: stop
usage: Usage(completion_tokens=238, prompt_tokens=98, total_tokens=336, completion_tokens_details=CompletionTokensDetailsWrapper(accepted_prediction_tokens=None, audio_tokens=None, reasoning_tokens=0, rejected_prediction_tokens=None, text_tokens=238, image_tokens=None, video_tokens=None), prompt_tokens_details=PromptTokensDetailsWrapper(audio_tokens=None, cached_tokens=0, text_tokens=None, image_tokens=None, video_tokens=None, cache_creation_tokens=0, cache_creation_token_details=CacheCreationTokenDetails(ephemeral_5m_input_tokens=0, ephemeral_1h_input_tokens=0)), cache_creation_input_tokens=0, cache_read_input_tokens=0, inference_geo='not_available', speed=None)

mk_schema

Agent

Runner

call_llm_tool

mk_transfer_doc

with_model_nm

Runner

Full agent example (1)

Advanced features

Preprocessing

Narrative casting

mk_narrative_cast

msg_fragment

History filtering

filter_out_history

Preproc

Agent.preproc_msg

Callbacks and outputkey

mk_cb_msg

tool_response

assistant_response

Agent.run_callback

Agent.maybe_save_to_output_key

Agent.__call__

Full agent example (2)

Sequential

Sequential

Agent.call