Daytona provider¶

The Daytona adapter executes code in a remote development environment. In the current MemoRizz adapter, each operation creates and removes its own environment. Consequently, sandbox_write_file followed by sandbox_read_file is not a coherent round trip; use E2B when a stateful provider filesystem is required.

Install and configure¶

pip install "memorizz[sandbox-daytona]"
export DAYTONA_API_KEY="<secret>"

Optional deployment settings:

export DAYTONA_API_URL="https://app.daytona.io/api"
export DAYTONA_TARGET="us"

Usage¶

from memorizz import MemAgentBuilder

agent = (
    MemAgentBuilder()
    .with_llm(my_llm)
    .with_sandbox(
        {
            "provider": "daytona",
            "api_url": "https://app.daytona.io/api",
            "target": "us",
        }
    )
    .build()
)

Direct provider use:

from memorizz.sandbox.providers.daytona_provider import DaytonaSandboxProvider

provider = DaytonaSandboxProvider()
result = provider.execute_code("print(6 * 7)")
assert result.stdout == ["42"]

Each call provisions a Daytona environment, runs process.code_run, normalizes the output, and removes the environment in a finally block.

Configuration reference¶

Setting	Default	Meaning
`api_key`	`DAYTONA_API_KEY`	Daytona credential.
`api_url`	`DAYTONA_API_URL` or provider default	API endpoint.
`target`	`DAYTONA_TARGET` or `us`	Deployment target.

Provider capabilities, limits, regions, and pricing can change independently of MemoRizz. Validate them against the Daytona deployment you operate.