The Start here course created this research-demo project and committed its docs-image, funks, and dagclass results. A project is a directory whose .dml/ state contains the local database, checkout, and project configuration. Use the CLI to administer it; Python opens the existing project to author or read research.
The documentation fixture owns a disposable Moto S3 endpoint. Configure that endpoint as this project’s single synchronization, script-execution, cache, and artifact endpoint. The assertions make the durable starting state explicit.
dml status | jq -e '.branch == "main"'
dml show | jq -e '.dags["docs-image"] and .dags.funks and .dags["number-pipeline"]'
dml config set remote.root "$DML_REMOTE_ROOT" >/dev/null
dml config show | jq -e --arg root "$DML_REMOTE_ROOT" '.remote.root == $root'true
true
trueConfiguration precedence is defaults, global configuration, project .dml/config.json, environment variables, then explicit CLI or Python overrides. Important project settings include default.branch_name, default.db_map_size_headroom, default.db_map_size_max, remote.root, remote.fetch_workers, and user. remote.root is not a dependency registry: add an import-only endpoint later with dml dep add NAME ROOT.
Python’s default session opens this initialized project. Dml(project_home=...) opens another existing project; use dml init rather than Dml.init() for the normal researcher workflow.
import daggerml as dml
assert dml.load("funks").result["first"].value() == {"count": 3, "total": 10}
with dml.temporary() as temporary:
with dml.new("scratch", dml=temporary) as dag:
result = dag.put("ephemeral", name="note")
dag.commit(result)
assert dml.load("scratch", dml=temporary).result.value() == "ephemeral"dml.temporary() creates an initialized isolated project and removes its state when the context exits. Use it for experiments and tests, not results that must be retained or shared. Continue with Artifacts, which records a durable input in this configured project.