Consolidate 'vertex_ai_proxy' UP usage and ensure service mapping for relevant UPs
Everyone can contribute. Help move this issue forward while earning points, leveling up and collecting rewards.
To proceed with the Refactor AiGateway.headers to accept unit_primitive and use pass it to Tokens.get issue, we need to ensure that every Unit Primitive that is explicitly used from permission checks has corresponding service with the same name (either defined explicitly in CC library config or virtually through AvailableServicesGenerator).
We tackle this in steps.
This issue is dedicated to UPs used in vertex_ai_proxy service/endpoint. Anthropic issue is here.
Currently, vertex_ai_proxy service lists these UPs:
| UP name | In use? | References in code/logs | Action |
|---|---|---|---|
|
|
|
|
Update owner in UP YAML |
|
|
|
|
Confirmed that this can be removed as it is not used, and there is an open MR to remove the client code. |
|
|
|
|
No action |
|
|
|
|
Remove from |
|
|
|
|
Check with code owner if this can be deleted from |
We need to carefully review the usage of the listed UPs and for each of them:
- Either remove it entirely from the linked service yml definition
- Or, if they are still used, remove them from https://gitlab.com/gitlab-org/cloud-connector/gitlab-cloud-connector/-/blob/main/src/ruby/lib/gitlab/cloud_connector/available_services_generator.rb#L11
IGNORE_UNIT_PRIMITIVESto make sure we have 1:1 mapping with the service
We should grep the codebase and use the logs from AI GW - https://log.gprd.gitlab.net/app/r/s/rdiwc - to understand the usage.
For safety, we need to wait until 18.2 final version is released before bumping the library (as it can be picked up by renovate and we can introduce the regression if there is a problem with the change).
We will use the table to receive approval from stakeholders and for transparency during the rollout.