Project definition files

A project definition file called snowflake.yml declares a directory as a Snowflake Native App project. It is a version-controlled file that resides at the root of a Snowflake Native App project directory and can either be created manually or by Snowflake CLI as part of project initialization. As long as you can provide this structured file in the directory but choose to use your own independent project structure, Snowflake CLI can discover the relevant files and carry out its functionality as usual.

For Native Apps, your snowflake.yml would look similar to the following:

definition_version: 2
entities:
  pkg:
    type: application package
    identifier: <name_of_app_pkg>
    stage: app_src.stage
    manifest: app/manifest.yml
    artifacts:
      - src: app/*
        dest: ./
      - src: src/module-add/target/add-1.0-SNAPSHOT.jar
        dest: module-add/add-1.0-SNAPSHOT.jar
      - src: src/module-ui/src/*
        dest: streamlit/
    meta:
      role: <your_app_pkg_owner_role>
      warehouse: <your_app_pkg_warehouse>
      post_deploy:
        - sql_script: scripts/any-provider-setup.sql
        - sql_script: scripts/shared-content.sql
  app:
    type: application
    identifier: <name_of_app>
    from:
      target: pkg
    debug: <true|false>
    meta:
      role: <your_app_owner_role>
      warehouse: <your_app_warehouse>
Copy

Common entity properties

The following table describes common properties available for project definition entities for Native Apps. See Specify entities for more information on project definition entities.

Common entity properties

Property

Definition

type

required, string

The type of entity to manage. For Snowflake Native App, valid values include:

identifier

optional, string

Optional Snowflake identifier for the entity, both unquoted and quoted identifiers are supported. To use quoted identifiers, include the surrounding quotes in the YAML value (e.g. ’”My Native Application Package”’).

If not specified, the entity ID in the project definition is used as the identifier.

meta.warehouse

optional, string

Warehouse used to run the scripts provided as part of meta.post_deploy, if any SQL commands within these scripts require use of warehouse.

Default: Warehouse specified for the connection in the Snowflake CLI config.toml file.

Note

If you do not specify a warehouse, the application passes validation, but fails to install.

Typically, you specify this value in the snowflake.local.yml as described in Project definition overrides.

meta.role

optional, string

Role to use when creating the entity and provider-side objects.

Note

If you do not specify a role, Snowflake CLI attempts to use the default role assigned to your user in your Snowflake account.

Typically, you specify this value in the snowflake.local.yml as described in Project definition overrides.

Default: Role specified in the Snowflake CLI connection

meta.post_deploy

optional, sequence

List of SQL scripts to execute after the entity has been created. The following example shows how to define these scripts in the project definition file:

definition_version: 2
entities:
  myapp_pkg:
    type: application package
    ...
    meta:
      post_deploy:
        - sql_script: scripts/post_deploy1.sql
        - sql_script: scripts/post_deploy2.sql
Copy

These scripts are invoked by commands that create or update an entity. For example, running the snow app deploy command executes these scripts after creating or updating a package. They are also executed by snow app run if the application instance is not being directly installed from a version or release directive.

You can also use templates in the post-deploy SQL scripts as well, as shown in the following sample script content:

GRANT reference_usage on database provider_data to share in entity <% fn.str_to_id(ctx.entities.myapp_pkg.identifier) %>
Copy

meta.use_mixins

optional, sequence

Names of mixins to apply to this entity. See Project mixins for more information

Application package entity properties

The following table describes common properties available for application package entities for Native Apps. See Specify entities for more information on project definition entities.

Properties for application package entities

Property

Definition

type

required, string

Must be application package.

manifest

required, string

The location of the Snowflake Native App manifest.yml file in your project.

deploy_root

optional, string

Subdirectory at the root of your project where the build step copies the artifacts. Once copied to this location, you can deploy them to a Snowflake stage.

Default: output/deploy

generated_root

optional, string

Subdirectory of the deploy root where Snowflake CLI writes generated files.

Default: __generated

stage

optional, string

Identifier of the stage that stores the application artifacts. The value uses the form <schema_name>.<stage_name>. The stage lives within the Application Package object. You can change the name to avoid name collisions.

Default: app_src.stage

artifacts

required, sequence

List of file source and destination pairs to add to the deploy root, as well as an optional Snowpark annotation processor. You can use the following artifact properties”

  • src: Path to the code source file or files

  • dest: Path to the directory to deploy the artifacts.

    Destination paths that reference directories must end with a /. A glob pattern’s destination that does not end with a / results in an error. If omitted, dest defaults to the same string as src.

    You can also pass in a string for each item instead of a dict, in which case the value is treated as both src and dest.

  • processors: Name of the processor to use to process the src code files. Currently, the only value supported is snowpark. For more information about custom processing, see Automatic SQL code generation and the snow app bundle command.

If src refers to just one file (not a glob), dest can refer to a target <path> or a <path/name>.

You can also pass in a string for each item instead of a dict, which case, the value is treated as both src and dest.

Example without a processor:

pkg:
  artifacts:
    - src: app/*
      dest: ./
    - src: streamlit/*
      dest: streamlit/
    - src: src/resources/images/snowflake.png
      dest: streamlit/
Copy

Example with a processor:

pkg:
  artifacts:
    - src: qpp/*
      dest: ./
      processors:
          - name: snowpark
            properties:
              env:
                type: conda
                name: <conda_name>
Copy

distribution

optional, string

Distribution of the application package created by the Snowflake CLI. When running snow app commands, Snowflake CLI warns you if the application package you are working with has a different value for distribution than is set in your resolved project definition.

Default: Internal

scratch_stage

optional, string

Identifier of the stage that stores temporary scratch data used by Snowflake CLI. The value uses the form <schema_name>.<stage_name>. The stage lives within the Application Package object. You can change the name to avoid name collisions.

Default: app_src.stage_snowflake_cli_scratch

Application entity properties

The following table describes common properties available for application entities for Native Apps. See Specify entities for more information on project definition entities.

Properties for application entities

Property

Definition

type

required, string

Must be application.

from.target

required, string

Application package from which to create this application entity. In the following example, target defines the name of an entity in the snowflake.yml file.

from:
  target: my_pkg
Copy

debug

optional, boolean

Whether to enable debug mode when using a named stage to create an application.

Default: True

More information about the artifacts.processors parameter

If you include the artifacts.processors parameter in the project definition file, the snow app bundle command invokes custom processing for Python code files in the src directory or file. Currently, Snowflake CLI supports only one processor, snowpark, which applies Snowpark annotation processing to Python files. The following shows the basic structure and syntax different processing environments:

  • To execute code in a conda environment, use the following:

    pkg:
      artifacts:
        - src: <some_src>
          dest: <some_dest>
          processors:
              - name: snowpark
                properties:
                  env:
                    type: conda
                    name: <conda_name>
    
    Copy

    where <conda_name> is the name of the conda environment containing the Python interpreter and the Snowpark library you want to use for Snowpark annotation processing.

  • To execute code in a Python virtual environment, use the following:

    pkg:
      artifacts:
        - src: <some_src>
          dest: <some_dest>
          processors:
              - name: snowpark
                properties:
                  env:
                    type: venv
                    path: <venv_path>
    
    Copy

    where <venv_path> is the path of the Python virtual environment containing the Python interpreter and the Snowpark library you want to use for Snowpark annotation processing. The path can be absolute or relative to the project directory.

  • To execute code in the currently active environment, use any of the following equivalent definitions:

    pkg:
      artifacts:
        - src: <some_src>
          dest: <some_dest>
          processors:
              - name: snowpark
                properties:
                  env:
                    type: current
    
    Copy

    or

    pkg:
      artifacts:
        - src: <some_src>
          dest: <some_dest>
          processors:
              - name: snowpark
    
    Copy

    or

    pkg:
      artifacts:
        - src: <some_src>
          dest: <some_dest>
          processors:
              - snowpark
    
    Copy

Project definition overrides

Though your project directory must have a snowflake.yml file, you can choose to customize the behavior of the Snowflake CLI by providing local overrides to snowflake.yml, such as a new role to test out your own application package. These overrides must be put in the snowflake.local.yml file that lives beside the base project definition. Snowflake suggests that you add it to your .gitignore file so it won’t be version-controlled by git. All templates provided by Snowflake already include it in the .gitignore file.

This overrides file must live in the same location as your snowflake.yml file.

The snowflake.local.yml file shares the exact schema as snowflake.yml, except that every value that was required is now optional, in additional to the already optional ones. The following shows a sample snowflake.local.yml file:

entities:
  pkg:
    meta:
      role: <your_app_pkg_owner_role>
      name: <name_of_app_pkg>
      warehouse: <your_app_pkg_warehouse>
  app:
    debug: <true|false>
    meta:
      role: <your_app_owner_role>
      name: <name_of_app>
      warehouse: <your_app_warehouse>
Copy

Every snow app command prioritizes the parameters in this file over those set in base snowflake.yml configuration file. Sensible defaults already provide isolation between developers using the same Snowflake account to develop the same application project, so if you are just getting started we suggest not including an overrides file.

The final definition schema obtained after overriding snowflake.yml with snowflake.local.yml is called the resolved project definition.

Limitations

Currently, Snowflake CLI does not support

  • Multiple override files.

  • A blank override file. Only create this file if you want to override a value from snowflake.yml.