dantalabs 0.2.0

DantaLabs SDK collection (featuring Maestro)

DantaLabs SDK

The DantaLabs SDK bundles the dlm command line tool and the MaestroClient Python library so that you can deploy, operate, and automate agents on the Maestro platform from one package.

Documentation

• dantalabs/docs/content/introduction.mdx ships with the package and mirrors the public documentation entry point.
• dantalabs/cli contains the Typer application behind dlm and is the best place to explore command behaviour.
• dantalabs/maestro/client.py exposes the Python client surface area, including resource helpers, service deployment utilities, and managed memory support.

Installation

pip install --upgrade dantalabs

# latest development build
pip install git+https://github.com/DantaLabs/maestro-sdk.git

Python 3.8 or newer is required.

Quick Start

1. Configure credentials

dlm setup
dlm status

# non-interactive environments
export MAESTRO_API_URL="https://dantalabs.com"
export MAESTRO_ORGANIZATION_ID="<org-id>"
export MAESTRO_AUTH_TOKEN="<token>"

You can override any stored value per command with --url, --org-id, or --token.

2. Deploy through the unified pipeline

dlm deploy ./agents/document-worker \
  --name "document-worker" \
  --agent-type script \
  --service

dlm deploy packages your project into a temporary bundle, uploads it, and optionally deploys the resulting agent as a managed Knative service. Use:

• --no-service to keep the definition without rolling out a service
• --definition-only to skip creating an agent instance
• --schema-file path/to/schema.json to attach JSON schema metadata

3. Call Maestro from Python

from dantalabs.maestro import MaestroClient

client = MaestroClient(
    organization_id="<org-id>",
    base_url="https://dantalabs.com",
    token="<token>",
)

agents = client.list_agents()
result = client.execute_agent_code_sync({"message": "Hello"}, agent_id=agents[0].id)

Managed memory, agent databases, file uploads, and network generation helpers are available through the same client instance.

CLI Highlights

• Essentials: dlm setup, dlm status, dlm version, dlm set-url
• Agents: dlm list-agents, dlm list-definitions, dlm create-agent, dlm update-agent, dlm use-agent, dlm run-agent
• Services: dlm service deploy, dlm service deployment-status, dlm service logs, dlm service execute, dlm service proxy
• Managed databases: dlm agentdb list, dlm agentdb inspect --show-connection, dlm agentdb connect --print-only
• Starters: dlm init clones the latest templates and dlm list-templates enumerates what is available

Every command accepts the standard authentication overrides (--org-id, --url, --token).

Python Client Highlights

• One MaestroClient instance exposes agents, executions, networks, files, and utils resource managers (dantalabs/maestro/resources).
• execute_agent_code(...) and execute_agent_code_sync(...) run script agents directly; deploy_service(...) launches Knative services programmatically.
• client.list_agent_databases(...), client.get_database_connection_info(...), and other helpers surface the managed PostgreSQL databases that back each agent.
• client.get_managed_memory(...) returns a dictionary-like object that persists state back to Maestro automatically when auto_save=True.
• File uploads (upload_file) and network scaffolding (generate_network) are available for enriching agent workflows.

Legacy Bundle Utilities (deprecated)

The CLI will continue to ship legacy bundle helpers for a short period to ease migration, but the unified deployment pipeline replaces the old workflows:

• dlm create-bundle, dlm upload-bundle, dlm deploy-bundle, dlm update-bundle, and dlm download-definition-bundle
• MaestroClient.create_bundle, upload_agent_bundle, create_and_upload_bundle, and related helpers in dantalabs/maestro/bundles

Prefer dlm deploy (or the /api/v1/agents/deploy endpoint) for new automation.

Development

Local development requires:

• Python >= 3.8
• httpx, pydantic, typer, python-dotenv, PyYAML, tomli (Python < 3.11), psycopg

Clone the repository, install dependencies into a virtualenv, and run the CLI with python -m dantalabs.cli for local testing.