https://www.boundaryml.com logo
Vertex AI doesn't work when used from client regis...
# help
a
A normal client setup like this works fine:
Copy code
client<llm> VertexGeminiFlash {
  provider vertex-ai
  retry_policy SimpleConstant
  options {
    model "gemini-2.5-flash"
    location us-central1
    credentials env.GOOGLE_APPLICATION_CREDENTIALS
    generationConfig {
      temperature 0.5
      thinkingConfig {
        thinkingBudget 0
      }
    }
  }
}
but using vertex in registry e.g.
Copy code
const tempRegistry = new ClientRegistry();
tempRegistry.addLlmClient('GeminiConnectorClient', 'vertex-ai', {
    model: 'gemini-2.5-flash',
    location: 'us-central1',
    credentials: process.env.GOOGLE_APPLICATION_CREDENTIALS,
    gemerationConfig: {
        temperature: 0.5,
        thinkingConfig: {
            thinkingBudget: 0
        }
    }
});
tempRegistry.setPrimary('GeminiConnectorClient');
and then using it will produce an error:
Copy code
BamlError: BamlClientError: BamlClientHttpError: Request failed with status code: 403 Forbidden. {"error":{"code":403,"message":"Method doesn't allow unregistered callers (callers without established identity). Please use API Key or other form of API consumer identity to call this API.","status":"PERMISSION_DENIED"}}
b
Yikes, thanks for flagging this, looking...
The second one does appear to have a typo -
gemerationConfig
should be an n, not an m
a
nevermind! this is fine actually
yeah besides that typo
b
Which version is this on?
a
false alarm it works totally fine, i was passing in the wrong client