Feature request: allow bundle init --config-file to read from Unity Catalog Volumes

Start at the bundle init command's --config-file handling and compare its local-file path with the existing Databricks-authenticated filesystem access used for Volume files. Verify that exact dbfs:/Volumes/ paths use the selected profile, preserve local-file behavior, avoid a local copy, and report authentication, permission, path, and JSON validation failures clearly.

Summary

Allow databricks bundle init to read its --config-file directly from an exact Unity Catalog Volume path.

For example:

databricks bundle init https://github.com/my-org/my-template \
  --config-file dbfs:/Volumes/customer_shared/config/bootstrap/customer-context.json \
  --output-dir ./my-solution

Motivation

Unity Catalog Volumes are a suitable governed location for non-secret operational configuration shared across customer workspaces.

Today, bundle init --config-file requires a local JSON file. A customer-owned bootstrap workflow must therefore:

1. authenticate to the workspace;
2. copy the exact file from the Volume to a temporary local file;
3. invoke bundle init;
4. remove the temporary file.

Direct Volume support would remove this wrapper and make governed, customer-owned bootstrap flows simpler while retaining Unity Catalog permissions and auditability.

Databricks already documents Unity Catalog Volumes as an appropriate location for configuration required across workspaces:

https://learn.microsoft.com/en-us/azure/databricks/files/files-recommendations

Proposed behavior

When --config-file receives an exact path beginning with dbfs:/Volumes/:

• read that single file using the existing Databricks authentication and selected profile;
• apply the same JSON parsing and template validation used for local files;
• do not scan the Volume or persist a local copy;
• fail clearly when authentication, permissions, the file path, or JSON validation fails.

Existing local-file behavior should remain unchanged.

Supporting /Volumes/... as a normalized alias would also be useful, although one canonical CLI form is sufficient.

Current workaround

databricks fs cp \
  dbfs:/Volumes/customer_shared/config/bootstrap/customer-context.json \
  /tmp/customer-context.json

databricks bundle init https://github.com/my-org/my-template \
  --config-file /tmp/customer-context.json \
  --output-dir ./my-solution

rm /tmp/customer-context.json

Alternative

If direct Unity Catalog file support is outside the intended scope of bundle init, supporting stdin would remove the temporary-file lifecycle:

databricks fs cat dbfs:/Volumes/.../customer-context.json \
  | databricks bundle init <template> --config-file -

Environment

Databricks CLI v1.11.0