Skip to content

Conversation

@andreynering
Copy link
Member

@andreynering andreynering commented Sep 17, 2025

This PR adds support for Google Vertex AI.

We're using the Google SDK for Gemini models and the Anthropic SDKs for Claude models. I had to do fixes on both to make it work properly for us.

While those are in review, I pushed the changes on an extra fantasy branch so we can import from this project.

On testing, ensure to use us-east5 region, because not all regions support Anthropic calls. It returns a generic 500 error if you try a not supported region.

@andreynering andreynering self-assigned this Sep 17, 2025
@andreynering andreynering force-pushed the charm-602-google-vertex-ai-initial-implementation branch 5 times, most recently from 1e8faaf to 28a8749 Compare September 24, 2025 14:05
@andreynering andreynering force-pushed the charm-602-google-vertex-ai-initial-implementation branch 2 times, most recently from 4b89e1b to 106ee37 Compare September 26, 2025 18:02
@andreynering andreynering force-pushed the charm-602-google-vertex-ai-initial-implementation branch 10 times, most recently from 0574c08 to 08840a9 Compare October 10, 2025 19:06
@andreynering andreynering marked this pull request as ready for review October 10, 2025 19:11
@andreynering andreynering force-pushed the charm-602-google-vertex-ai-initial-implementation branch from 08840a9 to 036fa3f Compare October 10, 2025 19:12
@andreynering andreynering changed the title WIP: feat: add google vertex support feat: add google vertex support Oct 10, 2025
@andreynering andreynering force-pushed the charm-602-google-vertex-ai-initial-implementation branch from 036fa3f to f69082a Compare October 10, 2025 19:24
Project: g.options.project,
Location: g.options.location,
}
if g.options.skipAuth {
Copy link
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

why is this needed?

Copy link
Member Author

@andreynering andreynering Oct 13, 2025

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

When running tests, specially when gcloud is not installed or configured, we need to skip authorization completely, because otherwise the internal SDKs will try to authenticate with gcloud and fail, and the tests themselves will fail. That means that the tests would always fail on CI, for example.

You made me realize that we might need to do a small adjustment here: we do need authorization to happen when recording, but we still need it to skip when just reproducing the VCRs.

Copy link
Member Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Just pushed a commit with a fix.

@andreynering andreynering enabled auto-merge (rebase) October 14, 2025 13:55
@andreynering andreynering force-pushed the charm-602-google-vertex-ai-initial-implementation branch from 482cf12 to 1b8c6c6 Compare October 14, 2025 13:57
@andreynering andreynering merged commit 83f044b into main Oct 14, 2025
16 checks passed
@andreynering andreynering deleted the charm-602-google-vertex-ai-initial-implementation branch October 14, 2025 14:01
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Projects

None yet

Development

Successfully merging this pull request may close these issues.

4 participants