googleapis / googleapis/python-aiplatform

Gemini controlled generation no longer supports schema fields that previously worked

Open
#4,551 0 comments 3 reactions 0 assignees View on GitHub
api: vertex-ai
Dominant language
Python
Stars
905
Forks
465
Avg merge
1d 13h
Merged PRs (30d)
44

Description

I am using Gemini's controlled generation function in one of my projects.
I use an OpenAPI schema with Vertex AI fields.

For one attribute, I had the following definition:
`{"type": "string", "format": "date", "example": "2015-12-04"}` which worked a few weeks ago, but currently results in:
`InvalidArgument : 400 Unable to submit request because one or more response schemas specified unsupported field example.`

If I remove the fields `format` and `example`, the whole generation works.

I cannot find release notes that would describe a change in this anywhere, did the support for this change? Both fields are specified in the Vertex AI Schema docs: https://cloud.google.com/vertex-ai/docs/reference/rest/v1/Schema,
but on the Controlled Generation page, they are largely limited: http://cloud.google.com/vertex-ai/generative-ai/docs/multimodal/control-generated-output#considerations

Contributor guide

Open the contributing guide

Research direction

Start by reproducing the reported controlled-generation request with the schema containing format and example. Read the linked Vertex AI Schema reference and Controlled Generation considerations, then determine whether the fields are supported and what compatibility or documentation change is expected.

Written by the indexing model from the issue text.

Assessment

Tech stack
google-cloud, python
Domain
ai, cloud
Issue type
Bug
Difficulty
3/5
Estimated time
1-2 days
Activity status
Stale
Clarity
Needs clarification
Newbie friendliness
35/100

Get new issues in your inbox

A short digest of beginner-friendly GitHub issues.