GoogleGenerativeAI Error [400 Bad Request]: contents.parts must not be empty

Questions
,
1

Describe the problem/error/question

I’m encountering an error in n8n when using my fine-tuned Gemini 1.5 flash model. Yesterday the setup was working perfectly, but today it’s throwing an error.

I’m also interested in integrating a Vertex AI fine-tuned Gemini 1.5 Pro model if that’s possible within n8n. Any information on that integration would be greatly appreciated.

What is the error message (if any)?

[GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/tunedModels/finetunelinkedin-hve38iyfflh4:streamGenerateContent?alt=sse: [400 Bad Request] * GenerateContentRequest.system_instruction.parts: contents.parts must not be empty.

Share the output returned by the last node

Information on your n8n setup

  • n8n version:
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app):
  • Operating system:
2

It looks like your topic is missing some important information. Could you provide the following if applicable.

  • n8n version:
  • Database (default: SQLite):
  • n8n EXECUTIONS_PROCESS setting (default: own, main):
  • Running n8n via (Docker, npm, n8n cloud, desktop app):
  • Operating system:
3
4

I have the same issue. No clue how to fix it unfortunately.
Google searches tell me it may be a node version issue.

5

same here.

1 Like
6

you need change existing value to $fromAI like this {{ $fromAI(‘start_date’) }}

here’s an example from google calendar API

image
image816×1183 59.4 KB

7

Hello,
The same issue occurred when I added the AI agent along with additional AI nodes.

8

Same issue here

9

bumping this

10

Any news on this? Is anyone also having troubles when calling tools with tools agent using Gemini?

Thanks.

11

same issue

image
image1186×575 43 KB

1 Like
12

Anyone from the n8n team have any word on this? Gemini is a great model for n8n and would like to get it working. Currently as a workaround im using openrouter with gemini, which has been really good with testing. And still free

13

Hi,
I’ve fixed it by adding a node.js node before.

I’ve just took data from the previous nodes to exit the same data :wink: . Maybe not the best answer but it works fine.

1 Like
14

I also started having this same error today with Gemini 2.0-flash-lite, it just stopped working and throwing this error out of nothing. Plugin in 4o-mini it works with no problem

n8n 1.82.0
self hosted docker
windows 10

15

I had the same problem and found the solution:

Your tool nodes’ names must not use special characters (such as ‘_’)!

16

not helps in mys case (deletel _ in tools - still - contents.parts must not be empty.

17

Same. I have added a node code before AI Agent node (output and input are the same) and it fixed.

1 Like
18

I´ve exactly the same issue with Google Gemini on an AI Agent. Also without connected tools the same problem arise. When I use Gemini on a simple AI Chain it works…

Here´s the exception: [GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-1.0-pro:streamGenerateContent?alt=sse: [400 Bad Request] * GenerateContentRequest.contents[1].parts: contents.parts must not be empty.

Maybe someone has another hint?

Best regards,
Michael

19

I found out that the reason for that seems to be a merge node before the agent node. As soon as I remove the merge node, also googles models work…

20

Hi Mike, I’m facing the same issue with Google Gemini on the AI Agent. I tried removing the Merge node before the AI Agent, but it didn’t help—the flow still gives the same error.