Feature: DakeraMemoryDriver — persistent cross-run memory backend for Griptape agents

[ferhimedamine]

Griptape has excellent in-session memory (ConversationMemory, TaskMemory) but no cross-run persistence. This proposes DakeraMemoryDriver as a persistent backend for Griptape agents using the self-hosted Dakera (https://dakera.ai) memory server.

Problem: Griptape agents lose all memory between process restarts. For agents running recurring tasks (daily report summaries, ongoing research assistants), context must be re-established on every run.

Proposed: DakeraMemoryDriver implementing BaseConversationMemoryDriver:

from griptape.drivers.memory.conversation import BaseConversationMemoryDriver
from griptape.artifacts import BaseArtifact
from dakera import DakeraClient

class DakeraConversationMemoryDriver(BaseConversationMemoryDriver):
base_url: str = 'http://localhost:3300'
api_key: str = ''
agent_id: str = 'griptape-agent'

def __post_init__(self):
    self._client = DakeraClient(base_url=self.base_url, api_key=self.api_key)

def store(self, memory: ConversationMemory) -> None:
    for run in memory.runs[-3:]:  # Store recent runs
        self._client.store_memory(
            agent_id=self.agent_id,
            content=f'Input: {run.input}\nOutput: {run.output}',
        )

def load(self) -> list:
    response = self._client.recall(agent_id=self.agent_id, query='recent conversations', top_k=10)
    return response.memories if response else []

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

agent = Agent(
    conversation_memory=ConversationMemory(
        driver=DakeraConversationMemoryDriver(
            base_url='http://localhost:3300',
            api_key='demo',
            agent_id='my-agent',
        )
    )
)

Setup: docker run -d -p 3300:3300 -e DAKERA_API_KEY=demo ghcr.io/dakera-ai/dakera:latest / pip install dakera

Relevant: griptape/drivers/memory/conversation/

Happy to open a PR.