Vertex AI Batch Mode, URL Context

Gemini API
,
1

Hi All,

I’m working as a small development team, using Gemini-2.5-flash via vertex AI. we’re currently working on batch processing some data, we’ve externally gathered URLs in our pipeline and we’re trying to feed these to Vertex for batches.

A single Gemini API call works (using Gemini on a single instance versus a JSONL batch), however we get the following error when trying the same payload with vertex.

Here’s a sample query:

{
  "request": {
    "contents": [
      {
        "parts": [
          {
            "text": "find me the latest news from apple https://www.apple.com/ "
          }
        ]
      }
    ],
    "tools": [
          {
              "url_context": {}
          }
      ]
    
  }
}

Below is the error payload.
'request.tools.url_context' with no fields at [1:1]

I’ve tried sifting through the api docs, and I’ve also looked at medium articles from April with similar payloads that worked historically but currently do not.

2 Likes
2

I also tried everything: Vertex AI, grounding, Google search. Then I tried the Google API in batch mode, but apparently the new SDK is failing.

The only way I get it to work is using the Gemini API without batch, so I easily hit the rate limit

3

Are you using python SDK? Maybe in JavaScript, Java, or Go works

5

I’m using a JSON payload and writing a JSONL file to a storage bucket to be run against vertex, unsure if since these features are experimental it’s not yet available in batch mode

7

Hello

Welcome to the Forum!!

Just to confirm, you are trying to use url_context tool in batch api call and you are getting error. I hope my understand is correct.

Would it be possible for you to share your file which you are using for batch call and your code, so that I can try to reproduce your issue to get more insights?