googleapis / googleapis/python-aiplatform
[Feature] CMEK support for Prompt Management with Vertex AI Python SDK
- Dominant language
- Python
- Stars
- 905
- Forks
- 465
- Avg merge
- 1d 13h
- Merged PRs (30d)
- 44
Description
It is not (yet) documented [documentation](https://cloud.google.com/vertex-ai/generative-ai/docs/learn/prompts/ai-powered-prompt-writing ) but from my test CMEK is not yet supported when using Prompt management with Vertex AI Python SDK. When using the UI, you can save prompt using CMEK.
Here what I saw when using 1.74.0 (don't see anything in the release note of the newer version)
Anyway this doesn't seems to work in my code [3] base on [2]:
```
Traceback (most recent call last):
File "/home/jupyter/proj_ml_at_scale_gcp_exp/src/gen-ai/vertex-ai-prompt-manangement.py", line 51, in
prompt1 = prompts.create_version(prompt=prompt)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jupyter/.conda-env/env_vertex_ai_llm/lib/python3.11/site-packages/vertexai/prompts/_prompt_management.py", line 298, in create_version
return _create_prompt_resource(prompt=prompt, version_name=version_name)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jupyter/.conda-env/env_vertex_ai_llm/lib/python3.11/site-packages/vertexai/prompts/_prompt_management.py", line 445, in _create_prompt_resource
dataset = _create_dataset(prompt=prompt, parent=parent)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jupyter/.conda-env/env_vertex_ai_llm/lib/python3.11/site-packages/vertexai/prompts/_prompt_management.py", line 393, in _create_dataset
operation = prompt._dataset_client.create_dataset(
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jupyter/.conda-env/env_vertex_ai_llm/lib/python3.11/site-packages/google/cloud/aiplatform_v1/services/dataset_service/client.py", line 943, in create_dataset
response = rpc(
^^^^
File "/home/jupyter/.conda-env/env_vertex_ai_llm/lib/python3.11/site-packages/google/api_core/gapic_v1/method.py", line 131, in __call__
return wrapped_func(*args, **kwargs)
^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
File "/home/jupyter/.conda-env/env_vertex_ai_llm/lib/python3.11/site-packages/google/api_core/grpc_helpers.py", line 78, in error_remapped_callable
raise exceptions.from_grpc_error(exc) from exc
google.api_core.exceptions.FailedPrecondition: 400 The request has violated one or more Org Policies: violations {
type: "constraints/gcp.restrictNonCmekServices"
subject: "orgpolicy:projects/PROJECT"
description: "Constraint `constraints/gcp.restrictNonCmekServices` violated for `projects/PROJECT` attempting to create a resource without specifying a KMS CryptoKey."
}
```
FR: add CMEK support for Prompt Management with Vertex AI python SDK (already supported when using the UI to save prompts). I know that this come later as the feature is still in preview but just to mention our interest as all regulated industry need such capability.
Contributor guide
Research direction
Start by tracing the prompt flow in vertexai/prompts/_prompt_management.py, especially _create_prompt_resource and _create_dataset, then inspect the DatasetService create_dataset request shown in the traceback. Determine how the SDK can pass a KMS CryptoKey when creating prompt resources. Done means prompt creation works under the stated restrictNonCmekServices policy, with coverage for the CMEK path and any needed documentation.
Written by the indexing model from the issue text.
Assessment
- Tech stack
- google-cloud, python
- Domain
- cloud, machine-learning, security
- Issue type
- Feature
- Difficulty
- 4/5
- Estimated time
- 3-5 days
- Activity status
- Stale
- Clarity
- Mostly clear
- Newbie friendliness
- 25/100