Files
humanlayer/examples/langchain/08-email-channel.py
2025-01-19 18:17:50 -08:00

117 lines
3.7 KiB
Python

from humanlayer import ContactChannel, EmailContactChannel
from langchain_core.prompts import ChatPromptTemplate
import langchain_core.tools as langchain_tools
from langchain_openai import ChatOpenAI
from langchain.agents import AgentExecutor, create_tool_calling_agent
from dotenv import load_dotenv
from humanlayer.core.approval import HumanLayer
load_dotenv()
hl = HumanLayer(
verbose=True,
contact_channel=ContactChannel(
email=EmailContactChannel(
address="dexter@humanlayer.dev",
context_about_user="the user you are helping",
)
),
# run_id is optional -it can be used to identify the agent in approval history
run_id="langchain-support-email",
)
task_prompt = """
You are the ceo's assistant, he contacts you via email with various tasks
The ceo has asked you to perform the following task:
----------------------------------
create a linear ticket for this issue and assign it to Austin
To: assistant@humanlayer.dev
From: ceo@humanlayer.dev
Subject: Fwd: Your annual Corportate report is ready
--
From: compliance@mosey.com
To: ceo@humanlayer.dev
Subject: Your annual Corportate report is ready
Hi Dexter,
Dear Dexter,
I hope this email finds you well. I am writing to inform you regarding your annual California Statement of Information (Form SI-550) submission requirement. Our records indicate that your report is now ready for filing with the California Secretary of State.
Please be advised that to avoid any late filing fees or penalties from the state of California, the submission deadline is November 31st. The current filing fee is $25, and late submissions may incur a $250 penalty. We strongly encourage you to complete this important California compliance requirement before the specified date.
Should you require any assistance with the filing process through the California Secretary of State's website or have any questions about California corporate filing requirements, please don't hesitate to reach out to our support team.
Best regards,
Compliance Team
Mosey Corporation
"""
def get_linear_projects() -> str:
"""get all linear projects"""
return [
{"id": "1", "name": "operations"},
{"id": "2", "name": "marketing"},
{"id": "3", "name": "engineering"},
{"id": "4", "name": "design"},
]
def get_linear_assignees() -> str:
"""get all linear assignees"""
return [
{"id": "1", "name": "Austin"},
{"id": "2", "name": "Dexter"},
]
@hl.require_approval()
def create_linear_ticket(title: str, assignee: str, description: str, project: str, due_date: str) -> str:
"""create a ticket in linear"""
if project != "operations":
return "that project does not exist"
return "issue created"
tools = [
langchain_tools.StructuredTool.from_function(get_linear_projects),
langchain_tools.StructuredTool.from_function(get_linear_assignees),
langchain_tools.StructuredTool.from_function(create_linear_ticket),
langchain_tools.StructuredTool.from_function(hl.human_as_tool()),
]
llm = ChatOpenAI(model="gpt-4o", temperature=0)
# Prompt for creating Tool Calling Agent
prompt = ChatPromptTemplate.from_messages(
[
(
"system",
"You are a helpful assistant.",
),
("placeholder", "{chat_history}"),
("human", "{input}"),
("placeholder", "{agent_scratchpad}"),
]
)
# Construct the Tool Calling Agent
agent = create_tool_calling_agent(llm, tools, prompt)
agent_executor = AgentExecutor(agent=agent, tools=tools, verbose=True)
if __name__ == "__main__":
result = agent_executor.invoke({"input": task_prompt})
print("\n\n----------Result----------\n\n")
print(result)