Conversation Memory

Adapted from the Griptape AI Framework documentation.

Overview

You can use Conversation Memory to give Gen AI Builder Structures the ability to keep track of the conversation across runs. All structures are created with ConversationMemory by default.

Example

Code
Logs

from griptape.structures import Agent

agent = Agent()

agent.run("My favorite animal is a Liger.")
agent.run("What is my favorite animal?")

[02/27/25 20:28:09] INFO     PromptTask acc8b8a0a736413791238b1d7c320d4d        
                             Input: My favorite animal is a Liger.              
[02/27/25 20:28:11] INFO     PromptTask acc8b8a0a736413791238b1d7c320d4d        
                             Output: That's interesting! Ligers are fascinating 
                             creatures. They are a hybrid offspring of a male   
                             lion and a female tiger. Ligers typically inherit  
                             physical and behavioral traits from both parents,  
                             often growing larger than either parent species.   
                             They are known for their unique appearance and are 
                             usually found in captivity, as lions and tigers do 
                             not share the same habitats in the wild. What do   
                             you find most interesting about ligers?            
                    INFO     PromptTask acc8b8a0a736413791238b1d7c320d4d        
                             Input: What is my favorite animal?                 
[02/27/25 20:28:12] INFO     PromptTask acc8b8a0a736413791238b1d7c320d4d        
                             Output: Your favorite animal is a liger.

You can disable conversation memory in any structure by setting it to None:

from griptape.structures import Agent

Agent(conversation_memory=None)

Interaction With Structures

Per Structure

By default, Conversation Memory Runs are created for each run of the structure. Gen AI Builder takes the Structure's input_task's input and the output_task's output, storing them in the Run. Tasks that are neither the input task nor the output task are not stored in the Run.

Code
Logs

from griptape.structures import Pipeline
from griptape.tasks import PromptTask

pipeline = Pipeline(
    tasks=[
        PromptTask("Respond to this request: {{ args[0] }}", id="input"),
        PromptTask("Improve the writing of this: {{ parent_output }}", id="improve"),
        PromptTask("Output this as a pirate: {{ parent_output }}", id="output"),
    ]
)

pipeline.run("My favorite animal is a Liger.")

if pipeline.conversation_memory is not None:
    for run in pipeline.conversation_memory.runs:
        print("Input", run.input.value)
        print("Output", run.output.value)
        print("\n\n")

[02/27/25 20:35:33] INFO PromptTask input
Input: Respond to this request: My favorite animal
is a Liger.
[02/27/25 20:35:34] INFO PromptTask input
Output: That's a unique choice! Ligers are
fascinating creatures, being hybrids between a male
lion and a female tiger. They are known for their
impressive size, often larger than both parent
species, and they exhibit traits from both lions
and tigers. While they don't occur naturally in the
wild, ligers are often found in captivity. What do
you find most interesting about ligers?
INFO PromptTask improve
Input: Improve the writing of this: That's a unique
choice! Ligers are fascinating creatures, being
hybrids between a male lion and a female tiger.
They are known for their impressive size, often
larger than both parent species, and they exhibit
traits from both lions and tigers. While they don't
occur naturally in the wild, ligers are often found
in captivity. What do you find most interesting
about ligers?
[02/27/25 20:35:36] INFO PromptTask improve
Output: That's an intriguing choice! Ligers are
captivating creatures, being hybrids of a male lion
and a female tiger. Known for their remarkable
size, they often surpass both parent species in
stature and display characteristics from both lions
and tigers. Although they don't occur naturally in
the wild, ligers are commonly found in captivity.
What aspect of ligers do you find most fascinating?
INFO PromptTask output
Input: Output this as a pirate: That's an
intriguing choice! Ligers are captivating
creatures, being hybrids of a male lion and a
female tiger. Known for their remarkable size, they
often surpass both parent species in stature and
display characteristics from both lions and tigers.
Although they don't occur naturally in the wild,
ligers are commonly found in captivity. What aspect
of ligers do you find most fascinating?
[02/27/25 20:35:37] INFO PromptTask output
Output: Arrr, that's a curious pick, matey! Ligers
be captivat'n beasts, they be, a mix of a lion
buccaneer and a tigress lass. Known fer their grand
size, they often outgrow both their lion and tiger
kin, showin' traits from both sides of the family.
Though ye won't find 'em roamin' the wild seas,
these mighty creatures be mostly found in
captivity. What part of these magnificent ligers be
sparklin' the most in yer eye, eh?
Input Respond to this request: My favorite animal is a Liger.
Output Arrr, that's a curious pick, matey! Ligers be captivat'n beasts, they be, a mix of a lion buccaneer and a tigress lass. Known fer their grand size, they often outgrow both their lion and tiger kin, showin' traits from both sides of the family. Though ye won't find 'em roamin' the wild seas, these mighty creatures be mostly found in captivity. What part of these magnificent ligers be sparklin' the most in yer eye, eh?

In this example, the improve Task is "forgotten" after the Structure's run is finished. This approach allows you to perform intermediary work within a Structure without it being stored in, and potentially cluttering, Conversation Memory.

Per Task

You can change when Conversation Memory Runs are created by modifying Structure.conversation_memory_strategy from the default per_structure to per_task.

Code
Logs

from griptape.structures import Pipeline
from griptape.tasks import PromptTask

pipeline = Pipeline(
    conversation_memory_strategy="per_task",
    tasks=[
        PromptTask("Respond to this request: {{ args[0] }}", id="input"),
        PromptTask("Improve the writing", id="improve"),
        PromptTask("Respond as a pirate", id="output"),
    ],
)

pipeline.run("My favorite animal is a Liger.")

if pipeline.conversation_memory is not None:
    for run in pipeline.conversation_memory.runs:
        print("Input", run.input.value)
        print("Output", run.output.value)
        print("\n\n")

[02/27/25 20:27:11] INFO PromptTask input
Input: Respond to this request: My favorite animal
is a Liger.
[02/27/25 20:27:15] INFO PromptTask input
Output: That's interesting! Ligers are fascinating
creatures, being hybrids between a male lion and a
female tiger. They are known for their impressive
size, often larger than both parent species, and
they exhibit a mix of traits from both lions and
tigers. While they don't occur naturally in the
wild, ligers are a testament to the diversity and
complexity of the animal kingdom. What do you find
most intriguing about ligers?
INFO PromptTask improve
Input: Improve the writing
[02/27/25 20:27:16] INFO PromptTask improve
Output: That's intriguing! Ligers are fascinating
creatures, being hybrids of a male lion and a
female tiger. Known for their impressive size, they
often grow larger than either parent species and
exhibit a blend of traits from both lions and
tigers. Although they don't occur naturally in the
wild, ligers showcase the diversity and complexity
of the animal kingdom. What do you find most
captivating about ligers?
INFO PromptTask output
Input: Respond as a pirate
[02/27/25 20:27:18] INFO PromptTask output
Output: Arrr, a liger, ye say? A mighty beast
indeed, a mix of lion and tiger, like a treasure
forged from the wildest parts of the animal
kingdom! They be as rare as a mermaid's tear, and
twice as fascinatin'. Tell me, matey, what be it
about these grand creatures that captures yer
adventurous spirit?
Input Respond to this request: My favorite animal is a Liger.
Output That's interesting! Ligers are fascinating creatures, being hybrids between a male lion and a female tiger. They are known for their impressive size, often larger than both parent species, and they exhibit a mix of traits from both lions and tigers. While they don't occur naturally in the wild, ligers are a testament to the diversity and complexity of the animal kingdom. What do you find most intriguing about ligers?

Input Improve the writing
Output That's intriguing! Ligers are fascinating creatures, being hybrids of a male lion and a female tiger. Known for their impressive size, they often grow larger than either parent species and exhibit a blend of traits from both lions and tigers. Although they don't occur naturally in the wild, ligers showcase the diversity and complexity of the animal kingdom. What do you find most captivating about ligers?

Input Respond as a pirate
Output Arrr, a liger, ye say? A mighty beast indeed, a mix of lion and tiger, like a treasure forged from the wildest parts of the animal kingdom! They be as rare as a mermaid's tear, and twice as fascinatin'. Tell me, matey, what be it about these grand creatures that captures yer adventurous spirit?

Now, each Task creates a Conversation Memory Run when it runs. This eliminates the need to feed the output of Tasks into each other using context variables like {{ parent_output }} since the output of the previous Task is stored in Conversation Memory and loaded when the next Task runs.

To blend the two approaches, you can disable Conversation Memory on individual tasks by setting PromptTask.conversation_memory to None.

Code
Logs

from griptape.structures import Pipeline
from griptape.tasks import PromptTask

pipeline = Pipeline(
    conversation_memory_strategy="per_task",
    tasks=[
        PromptTask("Respond to this request: {{ args[0] }}", id="input"),
        PromptTask("Improve the writing of this: {{ parent_output }}", id="improve", conversation_memory=None),
        PromptTask("Respond as a pirate", id="output"),
    ],
)

pipeline.run("My favorite animal is a Liger.")

if pipeline.conversation_memory is not None:
    for run in pipeline.conversation_memory.runs:
        print("Input", run.input.value)
        print("Output", run.output.value)
        print("\n\n")

[02/27/25 20:27:02] INFO PromptTask input
Input: Respond to this request: My favorite animal
is a Liger.
[02/27/25 20:27:03] INFO PromptTask input
Output: That's interesting! Ligers are fascinating
creatures, being hybrids between a male lion and a
female tiger. They are known for their impressive
size, often larger than both parent species, and
they exhibit a mix of traits from both lions and
tigers. While they don't occur naturally in the
wild, ligers have captured the imagination of many
due to their unique characteristics. What do you
find most interesting about ligers?
INFO PromptTask improve
Input: Improve the writing of this: That's
interesting! Ligers are fascinating creatures,
being hybrids between a male lion and a female
tiger. They are known for their impressive size,
often larger than both parent species, and they
exhibit a mix of traits from both lions and tigers.
While they don't occur naturally in the wild,
ligers have captured the imagination of many due to
their unique characteristics. What do you find most
interesting about ligers?
[02/27/25 20:27:04] INFO PromptTask improve
Output: That's intriguing! Ligers are captivating
creatures, resulting from the hybridization of a
male lion and a female tiger. Renowned for their
remarkable size, they often surpass both parent
species in stature and display a blend of traits
from lions and tigers. Although they do not occur
naturally in the wild, ligers have fascinated many
with their distinctive characteristics. What aspect
of ligers do you find most fascinating?
INFO PromptTask output
Input: Respond as a pirate
[02/27/25 20:27:06] INFO PromptTask output
Output: Arrr, a liger, ye say? A mighty beast
indeed, a mix of the king o' the jungle and the
striped hunter of the wild! 'Tis a creature as rare
as treasure on the high seas, with the strength of
a lion and the stealth of a tiger. Aye, if I were
to choose a beast to join me crew, a liger would
make a fine matey! What be it about the mighty
liger that captures yer fancy, me hearty?
Input Respond to this request: My favorite animal is a Liger.
Output That's interesting! Ligers are fascinating creatures, being hybrids between a male lion and a female tiger. They are known for their impressive size, often larger than both parent species, and they exhibit a mix of traits from both lions and tigers. While they don't occur naturally in the wild, ligers have captured the imagination of many due to their unique characteristics. What do you find most interesting about ligers?

Input Respond as a pirate
Output Arrr, a liger, ye say? A mighty beast indeed, a mix of the king o' the jungle and the striped hunter of the wild! 'Tis a creature as rare as treasure on the high seas, with the strength of a lion and the stealth of a tiger. Aye, if I were to choose a beast to join me crew, a liger would make a fine matey! What be it about the mighty liger that captures yer fancy, me hearty?

Types of Memory

Gen AI Builder provides several types of Conversation Memory to fit various use-cases.

Conversation Memory

ConversationMemory will keep track of the full task input and output for all runs.

Code
Logs

from griptape.memory.structure import ConversationMemory
from griptape.structures import Agent

agent = Agent(conversation_memory=ConversationMemory())

agent.run("Hello!")

print(agent.conversation_memory)

[02/27/25 20:27:28] INFO     PromptTask 3902de9d2ff5473e8f34398dfff5000e        
                             Input: Hello!                                      
                    INFO     PromptTask 3902de9d2ff5473e8f34398dfff5000e        
                             Output: Hello! How can I assist you today?         
{"type": "ConversationMemory", "runs": [{"type": "Run", "id": "9616891ea5eb4cb2a16d7eac246b7e7d", "meta": null, "input": {"type": "TextArtifact", "id": "9fa4bcf1f54840519d5fe027b4670565", "reference": null, "meta": {}, "name": "9fa4bcf1f54840519d5fe027b4670565", "value": "Hello!"}, "output": {"type": "TextArtifact", "id": "9a827d3528b14690b8cf5b1636e694b5", "reference": null, "meta": {"is_react_prompt": false}, "name": "9a827d3528b14690b8cf5b1636e694b5", "value": "Hello! How can I assist you today?"}}], "meta": {}, "max_runs": null}

You can set the max_runs parameter to limit how many runs are kept in memory.

Code
Logs

from griptape.memory.structure import ConversationMemory
from griptape.structures import Agent

conversation_memory = ConversationMemory(max_runs=2)
agent = Agent(conversation_memory=conversation_memory)

agent.run("Run 1")
agent.run("Run 2")
agent.run("Run 3")
agent.run("Run 4")
agent.run("Run 5")

print(conversation_memory.runs[0].input == "run4")
print(conversation_memory.runs[1].input == "run5")

[02/27/25 20:27:51] INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Input: Run 1                                       
[02/27/25 20:27:52] INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Output: Could you please provide more context or   
                             details about what you mean by "Run 1"? Are you    
                             referring to a specific event, a program, a race,  
                             or something else?                                 
                    INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Input: Run 2                                       
[02/27/25 20:27:53] INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Output: It seems like you're referencing something 
                             specific with "Run 1" and "Run 2," but without     
                             additional context, it's difficult to provide a    
                             meaningful response. Could you please provide more 
                             details or clarify what you're referring to?       
                    INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Input: Run 3                                       
[02/27/25 20:27:54] INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Output: It appears you're mentioning "Run 1," "Run 
                             2," and "Run 3," but without additional context,   
                             it's challenging to understand what you're         
                             referring to. Are these related to a sequence of   
                             events, a series of tests, a workout routine, or   
                             something else? Please provide more information so 
                             I can assist you better.                           
                    INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Input: Run 4                                       
[02/27/25 20:27:55] INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Output: It seems like you're referencing a sequence
                             or series, but without additional context, it's    
                             difficult to determine what "Run 4" specifically   
                             refers to. It could be related to a variety of     
                             things such as a series of experiments, a sequence 
                             in a game, a workout routine, or something else    
                             entirely. Could you provide more details or clarify
                             the context?                                       
                    INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Input: Run 5                                       
[02/27/25 20:27:56] INFO     PromptTask 2e9f9e5ff0664bd9882d14e9b20d2928        
                             Output: It seems like you're continuing a sequence 
                             with "Run 3," "Run 4," and now "Run 5." However,   
                             without additional context, it's difficult to      
                             determine what this sequence pertains to. Could you
                             provide more information or clarify the context so 
                             I can assist you more effectively?                 
False
False

Summary Conversation Memory

SummaryConversationMemory will progressively summarize task input and output of runs.

You can choose to offset which runs are summarized with the offset parameter.

Code
Logs

from griptape.memory.structure import SummaryConversationMemory
from griptape.structures import Agent
from griptape.utils import Conversation

agent = Agent(conversation_memory=SummaryConversationMemory(offset=2))

agent.run("Hello my name is John?")
agent.run("What is my name?")
agent.run("My favorite color is blue.")

print(Conversation(agent.conversation_memory))

[02/27/25 20:26:57] INFO PromptTask e40be7118727403a9fecb47655d7a0c4
Input: Hello my name is John?
INFO PromptTask e40be7118727403a9fecb47655d7a0c4
Output: Hello John! How can I assist you today?
INFO PromptTask e40be7118727403a9fecb47655d7a0c4
Input: What is my name?
[02/27/25 20:26:58] INFO PromptTask e40be7118727403a9fecb47655d7a0c4
Output: Your name is John. How can I help you
further?
INFO PromptTask e40be7118727403a9fecb47655d7a0c4
Input: My favorite color is blue.
[02/27/25 20:26:59] INFO PromptTask e40be7118727403a9fecb47655d7a0c4
Output: Blue is a great color! It's often
associated with calmness and serenity. Do you have
any other favorites or anything else you'd like to
share?
Q: Hello my name is John?
A: Hello John! How can I assist you today?
Q: What is my name?
A: Your name is John. How can I help you further?
Q: My favorite color is blue.
A: Blue is a great color! It's often associated with calmness and serenity. Do you have any other favorites or anything else you'd like to share?
Summary: John introduces himself to the assistant, who then offers to help him.

← Prev

Task Memory and Off Prompt

↑ Up

Agents

Rulesets

Could this page be better? Report a problem or suggest an addition!

Conversation Memory

Overview

Example

Interaction With Structures

Per Structure

Per Task

Types of Memory

Conversation Memory

Summary Conversation Memory

← Prev

↑ Up

Next →