External Inference using google vertex ai and gemini-3.6-flash error

Elastic Stack Elastic Agent
1

Hi!

I'm trying to set up an external inference using our Google vertex ai account and the model gemini-3.6-flash. The inference creation works, but trying to use it as a chat-completion fails with an error:

Error: Received an unsuccessful status code for request from inference
entity id [googlevertexai-chat_completion-example] status [400].

Function call is missing a thought_signature in functionCall parts.
This is required for tools to work correctly, and missing
thought_signature may lead to degraded model performance.

Additional data:
function call `default_api:platform_core_list_indices`, position 2.

We are on Elastic Cloud Hosted using version 9.5.2

2

Hey @tallakh. Unfortunately there's a compatibility limitation with Gemini 3. You'll need to use a Gemini 2.x model such as gemini-2.5-flash. I'll improve our public-facing documentation around this limitation this while the eng team looks at improving compatibility for newer models.

3

Thanks @bmorelli25 Hope this will be fixed soon, Gemini 2.x models feels a bit ancient :slight_smile:

5

Hey again. Wanted to let you know that a fix has been merged into Elasticsearch and will be released in an upcoming release. https://github.com/elastic/elasticsearch/pull/160167