Skip to content

Formatting Output

Rich Library

To make the chatbot output look more chat-like, we'll use the rich library. This library provides advanced formatting and styling options for the console output. We'll modify the chatbot function to apply formatting to the agent's responses.


First, let's update the code to import the rich library. Include the following import statements in the import section of

from rich import print as rprint
from rich.panel import Panel

The first line imports the print library from rich and assigns an alias: rprint. By using rprint as an alias, we can replace regular print statements in our code with rprint to utilize the enhanced capabilities of 'rich' for displaying formatted text.

For example, instead of using print("Hello, World!"), we can now use rprint("Hello, World!") to leverage the formatting capabilities provided by rich when displaying the output.


Sometimes people will simply recommend overriding the standard print functionality by doing from rich import print, but that would actually replace other uses of print in your code. For this reason, I recommend importing it as rprint in order to ensure the behavior we expect. But in reality, it's totally up to you. Read the documentation for more information.

The second line imports the Panel class from the rich.panel module. The Panel class represents a styled container that can be used to encapsulate and visually enhance content within a console output. It allows us to create panels with various styles, colors, and borders.


Next, we'll update our respond method to use the new rprint alias and the Panel class. This is a pretty simple change to start with, but you'll very quickly see how much nicer things look.

Inside the respond method, replace the line that looks like:

print(f"Kiwi: {response}")

rprint(Panel(f"Kiwi: {response}"))

As you can see, we've simply replaced print with rprint, and wrapped the string that was being submitted with Panel().

If you run this code you'll see a quick improvement.

│ Kiwi: Kia Ora! What can I do for you today?                                     │

Chat with Kiwi: 

Much better, right? We're not done yet...

Fitting it in

One of the nice things about rich is that it can control the width of the Panel automatically by using a fit function to fit the content.

Modify the Panel line to include .fit

rprint("Kiwi: {response}"))

Try it out to see how it feels.

Chat with Kiwi: Say hello in 2 words as a kiwi

│ Kiwi: Kia ora, mate! │

Proper width

Sometimes the response can be quite long and fill the terminal. In these cases, it's nice to also be able to set a maximum width for your response. You can do this by specifying the width parameter of When width is specified, the resulting panel will be either the width of your content or the width you specified - whatever is smaller.

Modify the prompt:

rprint("Kiwi: {response}", width=80))

Now the panel will be at most 80 characters wide.

Chat with Kiwi: What's the best thing about the Wairarapa?

│ Kiwi: Oh, the Wairarapa, mate! It's a stunner. The best thing about it has   │
│ to be the beautiful landscapes, from the rugged coastlines to the lush       │
│ vineyards. It's a real treat for the eyes, I tell ya!                        │

Code Review

As you can see, this has already helped our readability a ton. Compare your code.
from dotenv import load_dotenv
import logging
import json

# Rich
from rich import print as rprint
from rich.panel import Panel

# Griptape 
from griptape.structures import Agent
from griptape.rules import Rule, Ruleset

# Load environment variables

# Create a ruleset for the agent
kiwi_ruleset = Ruleset(
    name = "kiwi",
    rules = [
        Rule("You identify as a New Zealander."),
        Rule("You have a strong kiwi accent.")

json_ruleset = Ruleset(
        Rule("Respond in plain text only with JSON objects that have the following keys: response, continue_chatting."),
        Rule("The 'response' value should be a string that is your response to the user."),
        Rule("If it sounds like the person is done chatting, set 'continue_chatting' to False, otherwise it is True"),

# Create a subclass for the Agent
class MyAgent(Agent):

    def respond (self, user_input):
        agent_response =
        data = json.loads(agent_response.output_task.output.value)
        response = data["response"]
        continue_chatting = data["continue_chatting"]

        rprint("Kiwi: {response}", width=80))

        return continue_chatting

# Create the agent
agent = MyAgent(
    rulesets=[kiwi_ruleset, json_ruleset],

# Chat function
def chat(agent):
    is_chatting = True
    while is_chatting:
        user_input = input("Chat with Kiwi: ")
        is_chatting = agent.respond(user_input)

# Introduce the agent
agent.respond("Introduce yourself to the user.")

# Run the agent

Next Steps

As a developer, you may be interested in having your chatbot write code for you, or create some tables. In the next section: Markdown Madness, we'll take a look at the Markdown class in rich, and use it to ensure output looks as we expect.