Default environment variables

Read these variables in your code instead of hard-coding Project names, IDs, or file paths.

Use your own variables

Three kinds of variables carry your own values into executions. They differ in who owns the value and where it applies, so choose by what the value belongs to: a person, a Project, or an environment.

Kind	Applies to	Use it for
User variables	Every execution you start, in any Project	Credentials that belong to you
Project variables	Every execution in the Project, whoever starts it	Service account credentials and shared settings
Compute Environment variables	Every execution that uses the environment, in any Project	Tool configuration and values computed when an execution starts

User variables

User variables carry credentials that belong to one person, such as your database login, API tokens, or Git credentials. Domino injects them into every Workspace, Job, and App you start. Set a credential once, and every Project you work in picks it up. When a collaborator runs the same code, Domino injects their values instead of yours. Each person connects with their own identity and permissions, and the systems they connect to record who did what. A scheduled Job uses its owner’s user variables. An App uses the variables of the user who started it, so everyone who views the App acts with that user’s credentials. Give an App a service account through a Project variable instead. Reuse coding assistant credentials shows user variables and a pre-run script working together.

Project variables

Project variables carry values that belong to the work rather than to a person. Every execution in the Project receives them, whoever starts it. Store a service account’s credentials here, so scheduled Jobs, Apps, and Flows don’t depend on one person’s login. They keep running when the person who set them up is away or leaves the team. Project variables also keep shared settings consistent, such as a bucket name, an API base URL, or a model threshold. Change a value once, and every collaborator’s next execution uses it. A Project can also export its variables to Projects that import it, so one Project can supply configuration to several others. Anyone who can run code in the Project can read its variables. Give a service account only the access the Project needs.

Compute Environment variables

Compute Environment variables apply to every execution that uses the environment, in any Project. Set them in the environment definition in two ways:

An ENV instruction in the Dockerfile sets a fixed value when the image builds. Use it for configuration that every Project needs, such as a proxy address, a package index URL, or a cache directory.
A pre-run script exports values when each execution starts, so a value can depend on the execution. The script can read the default variables and your user and Project variables, and compute new values from them. It can add a Project folder to the Python path with export PYTHONPATH=$DOMINO_WORKING_DIR/src, size thread pools to the execution’s CPUs, or exchange a user’s credential for a short-lived token.

Don’t put secrets in ENV, because anyone who can use the image can read them. Use user or Project variables for credentials.

When names collide

When more than one source sets the same name, later sources win: Compute Environment ENV values, then Project variables, then user variables, then pre-run script exports. A Project can set a default, such as a shared read-only login, that one user replaces with a user variable of the same name.

Don’t name your own variables after the defaults on this page or with the DOMINO_ prefix. Domino doesn’t block such names, but a duplicate can override Domino’s value or be overwritten by it. See reserved variable names.

Project and execution

Variable	Value
`DOMINO_PROJECT_ID`	ID of the Project.
`DOMINO_PROJECT_NAME`	Name of the Project.
`DOMINO_PROJECT_OWNER`	Username of the Project owner.
`DOMINO_RUN_ID`	ID of the execution.
`DOMINO_RUN_NUMBER`	Number of the execution within the Project.
`DOMINO_STARTING_USERNAME`	Username of the user who started the execution.
`DOMINO_STARTING_USER_ID`	User ID of the user who started the execution.
`DOMINO_USER_NAME`	Username of the user who started the execution.
`DOMINO_HARDWARE_TIER_ID`	ID of the hardware tier the execution runs on.
`DOMINO_ENVIRONMENT_ID`	ID of the execution’s Compute Environment. Set in Workspaces, Apps, and Launcher runs.
`DOMINO_ENVIRONMENT_REVISION_ID`	ID of the Compute Environment revision. Set in Workspaces, Apps, and Launcher runs.
`DOMINO_IS_GIT_BASED`	`true` in a Git-based Project, `false` in a Domino File System (DFS) Project.
`DOMINO_IS_WORKFLOW_JOB`	`true` when a Flow started the Job, otherwise `false`.
`DOMINO_IS_LOCAL_DATA_PLANE`	`true` when the execution runs on the local Data Plane, `false` on a remote Data Plane.
`DOMINO_NODE_NAME`	Name of the compute node that runs the execution.
`DOMINO_NODE_IP`	IP address of that compute node.
`DOMINO_VERSION`	Domino version, such as `6.3.1`. Set in Workspaces when Workspace skills updates are on, the default.
`DOMINO_USER_ID`	`12574`, the Linux user ID that runs your code.
`DOMINO_GROUP_ID`	`12574`, the Linux group ID.

File paths

Paths depend on the Project type, and a variable missing from a Project type’s table isn’t set there. DOMINO_NETAPP_VOLUMES_DIR is set only in Workspaces, Apps, and Launcher runs that mount a NetApp Volume.

Git-based Projects

Variable	Path	Contents
`DOMINO_WORKING_DIR`	`/mnt/code`	Project code
`DOMINO_ARTIFACTS_DIR`	`/mnt/artifacts`	Project artifacts
`DOMINO_DATASETS_DIR`	`/mnt/data`	The Project’s Datasets
`DOMINO_IMPORTED_DATA_DIR`	`/mnt/imported/data`	Datasets from other Projects
`DOMINO_IMPORTED_ARTIFACTS_DIR`	`/mnt/imported/artifacts`	Artifacts from imported Projects
`DOMINO_IMPORTED_CODE_DIR`	`/mnt/imported/code`	Imported Git repositories
`DOMINO_NETAPP_VOLUMES_DIR`	`/mnt/netapp-volumes`	NetApp Volumes

DFS Projects

Variable	Path	Contents
`DOMINO_WORKING_DIR`	`/mnt`, or `/mnt/<owner>/<project>` when the Project imports another Project	Project files and artifacts
`DOMINO_DATASETS_DIR`	`/domino/datasets`	Datasets: the Project’s own under `local/`, others directly under this path
`DOMINO_IMPORTED_CODE_DIR`	`/repos`	Imported Git repositories
`DOMINO_IMPORTED_PROJECTS_DIR`	`/mnt`	Imported Projects
`DOMINO_NETAPP_VOLUMES_DIR`	`/domino/netapp-volumes`	NetApp Volumes

See Understand Dataset paths for snapshot paths.

Variables for both Project types

A Project that imports other Projects also gets DOMINO_<OWNER>_<PROJECT>_WORKING_DIR for each one, set to that Project’s mount path. Domino upper-cases the owner’s username and the Project name and replaces every character other than a letter or digit with _. A Project named churn-model owned by jane.doe sets DOMINO_JANE_DOE_CHURN_MODEL_WORKING_DIR. DOMINO_TRAINING_SET_PATH points to TrainingSet storage at /var/opt/domino/trainingset/project. Domino sets it in executions that mount a Dataset or can use the Project’s TrainingSets.

Domino services

Variable	Value
`DOMINO_API_PROXY`	Address of a local proxy that adds your access token to Domino API requests, such as `http://localhost:8899`. Send Domino API calls here; see Use the API proxy.
`DOMINO_USER_HOST`	Address of the Domino API from inside the execution. Use `DOMINO_API_PROXY` for authenticated calls.
`MLFLOW_TRACKING_URI`	Address of the Domino experiment tracking server. MLflow clients read it automatically.
`DOMINO_DATASOURCE_PROXY_HOST`	Address of the Data Source proxy. The Domino Data API reads it.
`DOMINO_DATASOURCE_PROXY_FLIGHT_HOST`	Arrow Flight address of the Data Source proxy. The Domino Data API reads it.
`DOMINO_DATA_API_GATEWAY`	Address the Domino Data API uses for vector database Data Sources.
`DOMINO_MLFLOW_DEPLOYMENTS`	Address of the MLflow Deployments server that AI Gateway (legacy) is built on.

Apps and Workspaces

Variable	Value
`DOMINO_RUN_HOST_PATH`	URL path that Domino’s reverse proxy routes to the Workspace or App, ending in `/`. Use it as your App’s base path; see Request headers and environment variables for Apps.
`DOMINO_APP_ID`	ID of the App. Set in deployed agents only.
`DOMINO_APP_VERSION_ID`	ID of the App version. Set in deployed agents only.
`DOMINO_AGENT_IS_PROD`	`true` in deployed agents. Puts `init_tracing()` in the Agents Python SDK into production mode.
`DOMINO_WORKSPACE_PERSISTS_HOME`	`true` in a Workspace with Home Directory Persistence on.

Deployed agents are Apps that you deploy as agentic systems.

Credentials

Variable	Value
`AWS_SHARED_CREDENTIALS_FILE`	Path to the Amazon Web Services (AWS) credentials file that Domino writes when AWS credential propagation is enabled. AWS client libraries and command-line tools read it automatically.
`AWS_CREDENTIAL_PROFILES_FILE`	The same path, set when AWS credential propagation is enabled and a Spark cluster is attached.
`KRB5_CONFIG`	Path to a `krb5.conf` file. Set when Kerberos authentication is configured and the Project, or a Project it imports, contains the file.

Both AWS variables point to /var/lib/domino/home/.aws/credentials. Domino sets them in Workspaces and Jobs, including Flow tasks, but not in Apps, Launcher runs, or scheduled Jobs.

Domino endpoints

A Domino endpoint serves a model and doesn’t receive run, user, or Project variables. Its image is built from the Compute Environment image, so the environment’s ENV values carry over. Domino also writes these variables into the image when it builds each version, replacing any ENV value with the same name:

Variable	Value
`DOMINO_PROJECT_NAME`	Name of the Project when the version was built.
`DOMINO_PROJECT_OWNER`	Username of the Project owner when the version was built.
`DOMINO_WORKING_DIR`	Location of the Project files: `/mnt/code` for a Git-based Project, `/mnt/<owner>/<project>` for a DFS Project.
`DOMINO_USER_HOST`	Internal address of the Domino API.
`DOMINO_DATASOURCE_PROXY_HOST`	Address of the Data Source proxy.
`DOMINO_DATASOURCE_PROXY_FLIGHT_HOST`	Arrow Flight address of the Data Source proxy.
`MLFLOW_MODEL_URI`	`$DOMINO_WORKING_DIR/mlflow`, which holds the registered model. Set only for versions built from the Model Registry.

Build a new version to pick up a Project rename or ownership change. To pass other values to an endpoint, set Domino endpoint variables.

Compute clusters

When you attach a cluster to a Workspace or Job, Domino adds variables that point your code at the cluster:

Variable	Value	Cluster
`SPARK_MASTER_HOST`, `SPARK_MASTER_PORT`	Host name and port (`7077`) of the Spark master.	Spark
`RAY_HEAD_SERVICE_HOST`, `RAY_HEAD_SERVICE_PORT`	Host name and port (`10001`) of the Ray head’s client service.	Ray
`DASK_SCHEDULER_SERVICE_HOST`, `DASK_SCHEDULER_SERVICE_PORT`	Host name and port (`8786`) of the Dask scheduler.	Dask
`OMPI_MCA_orte_default_hostfile`	`/domino/mpi/hosts`, the host file that lists the Message Passing Interface (MPI) workers.	MPI
`SLURM_CONF_SERVER`	Address of the Slurm controller.	Slurm
`SBATCH_PARTITION`	Slurm partition for the cluster.	Slurm
`CLUSTER_MLFLOW_TRACKING_URI`	Value of `MLFLOW_TRACKING_URI` inside the cluster’s pods.	Any

Code on the cluster’s own pods, such as Spark executors or Ray workers, runs in the cluster’s Compute Environment and receives only a few Domino variables, such as DOMINO_API_PROXY, DOMINO_USER_HOST, and MLFLOW_TRACKING_URI. It doesn’t receive Project, run, or user variables, or your own user and Project variables. Pass the values your cluster code needs from the Workspace or Job, for example as task arguments. A Slurm job that you submit with sbatch from a Workspace inherits the Workspace’s environment, because Slurm exports the submitting environment by default. High-performance computing (HPC) Jobs that Domino submits for you receive no Domino variables, only PATH and Slurm’s own job variables, such as SLURM_JOB_ID.

Deprecated variables

Domino keeps these for older code. Use the replacement in new code.

Variable	Value	Use instead
`DOMINO_API_HOST`	Same value as `DOMINO_USER_HOST`.	`DOMINO_USER_HOST`, or `DOMINO_API_PROXY` for authenticated calls
`DOMINO_USER_API_KEY`	API key of the user who started the execution, when that user has one.	`DOMINO_API_PROXY`

Manage environment variables explains how to set your own variables at the Compute Environment, Project, endpoint, and user levels.
Access environment variables shows how to read a variable in Python and R.
Domino API authentication covers the API proxy and token options.
Request headers and environment variables for Apps lists the headers Domino adds to App requests.

​Use your own variables

​User variables

​Project variables

​Compute Environment variables

​When names collide

​Project and execution

​File paths

​Git-based Projects

​DFS Projects

​Variables for both Project types

​Domino services

​Apps and Workspaces

​Credentials

​Domino endpoints

​Compute clusters

​Deprecated variables

​Related

Use your own variables

User variables

Project variables

Compute Environment variables

When names collide

Project and execution

File paths

Git-based Projects

DFS Projects

Variables for both Project types

Domino services

Apps and Workspaces

Credentials

Domino endpoints

Compute clusters

Deprecated variables

Related