How to Configure Agents Instructions#

Prerequisites

This guide assumes familiarity with Agents.

Agents can be configured to tackle many scenarios. Proper configuration of their instructions is essential.

In this how to guide, we will learn how to:

  • Configure the instructions of an Agent.

  • Set up instructions that vary with each Conversation.

  • Maintain instructions that are consistently updated and refreshed.

WayFlow supports several LLM API providers. Select an LLM from the options below:

from wayflowcore.models import OCIGenAIModel

if __name__ == "__main__":

    llm = OCIGenAIModel(
        model_id="provider.model-id",
        service_endpoint="https://url-to-service-endpoint.com",
        compartment_id="compartment-id",
        auth_type="API_KEY",
    )

Basic implementation#

Assuming you need an agent to assist a user in writing articles, use the implementation below for that purpose:

from wayflowcore.agent import Agent

agent = Agent(
    llm=llm,
    custom_instruction="""Your a helpful writing assistant. Answer the user's questions about article writing.
Make sure to welcome the user first, but keep it short""",
    initial_message=None,
)

Then execute it:

conversation = agent.start_conversation()
conversation.execute()
last_message = conversation.get_last_message()
print(last_message.content if last_message else "No message")
# Welcome to our article writing guide. How can I assist you today?

Sometimes, there is contextual information relevant to the conversation. Assume a user is interacting with the assistant named “Jerry.” To make the assistant more context-aware, define a variable or expression in the custom_instruction Jinja template, and pass it when creating the conversation:

agent = Agent(
    llm=llm,
    custom_instruction="""Your a helpful writing assistant. Answer the user's questions about article writing.
Make sure to welcome the user first, their name is {{user_name}}, but keep it short""",
    initial_message=None,
)

conversation = agent.start_conversation(inputs={"user_name": "Jerry"})
conversation.execute()
last_message = conversation.get_last_message()
print(last_message.content if last_message else "No message")
# Hello Jerry, I'm here to help with any article writing-related questions you may have. Go ahead and ask away!

Note

It is useful to use the same Agent, but change some part of the custom_instruction for each different conversation.

Finally, incorporating dynamic context into the agent’s instructions can significantly improve its responsiveness. For example, the instructions can contain the current time making the agent more aware of the situation. The time value is constantly changing, so you need to make sure it is always up-to-date. To do this, use the ContextProvider:

from datetime import datetime

from wayflowcore.contextproviders import ToolContextProvider
from wayflowcore.tools import tool

@tool
def get_current_time() -> str:
    """Tool that gets the current time"""
    return datetime.now().strftime("%d, %B %Y, %I:%M %p")

time_provider = ToolContextProvider(tool=get_current_time, output_name="current_time")

agent = Agent(
    llm=llm,
    custom_instruction="""Your a helpful writing assistant. Answer the user's questions about article writing.
It's currently {{current_time}}.""",
    initial_message=None,
    context_providers=[time_provider],
)

conversation = agent.start_conversation()
conversation.execute()
last_message = conversation.get_last_message()
print(last_message.content if last_message else "No message")
# I'm here to assist you with any article writing-related questions or concerns. What do you need help with today?

You successfully customized the prompt of your agent.

Recap#

In this guide, you learned how to configure Agent instructions with:

  • pure text instructions;

  • specific variables for each Conversation;

  • instructions with variables that needs to be always updated.

Below is the complete code from this guide.
from wayflowcore.models import LlmModelFactory

llm = LlmModelFactory.from_config(model_config)

from wayflowcore.agent import Agent

agent = Agent(
    llm=llm,
    custom_instruction="""Your a helpful writing assistant. Answer the user's questions about article writing.
Make sure to welcome the user first, but keep it short""",
    initial_message=None,
)

conversation = agent.start_conversation()
conversation.execute()
last_message = conversation.get_last_message()
print(last_message.content if last_message else "No message")
# Welcome to our article writing guide. How can I assist you today?

agent = Agent(
    llm=llm,
    custom_instruction="""Your a helpful writing assistant. Answer the user's questions about article writing.
Make sure to welcome the user first, their name is {{user_name}}, but keep it short""",
    initial_message=None,
)

conversation = agent.start_conversation(inputs={"user_name": "Jerry"})
conversation.execute()
last_message = conversation.get_last_message()
print(last_message.content if last_message else "No message")
# Hello Jerry, I'm here to help with any article writing-related questions you may have. Go ahead and ask away!

from datetime import datetime

from wayflowcore.contextproviders import ToolContextProvider
from wayflowcore.tools import tool

@tool
def get_current_time() -> str:
    """Tool that gets the current time"""
    return datetime.now().strftime("%d, %B %Y, %I:%M %p")

time_provider = ToolContextProvider(tool=get_current_time, output_name="current_time")

agent = Agent(
    llm=llm,
    custom_instruction="""Your a helpful writing assistant. Answer the user's questions about article writing.
It's currently {{current_time}}.""",
    initial_message=None,
    context_providers=[time_provider],
)

conversation = agent.start_conversation()
conversation.execute()
last_message = conversation.get_last_message()
print(last_message.content if last_message else "No message")
# I'm here to assist you with any article writing-related questions or concerns. What do you need help with today?

Next steps#

Having learned how to configure agent instructions, you may now proceed to: