AI - First Time configuration (How To)

[Felipereis50]

Hi friends,

For the first time, I’m trying to configure an AI.
I’m still learning, and I’m using GEMINI

First
First I asked in this forum if it would be possible to connect to Gemini.

Can I connect with the Gemini API? Is there a connector for Gemini?

Can I connect with the Gemini API? Is there a Node.js connector for Gemini?

Second
I downloaded this workflow and I want to change the configuration of the LLM to Gemini, and try to test this workflow with the Gemini API.

Sentiment Analysis - Generative AI – knime

My Steps

1-Credentials Configuration
I’ve got the API Key from Google Ai Studio,

I have copied and paste to the Credentials node.

2-Open AI Authenticator

• I’ve passed the variable "credential
• Then copy paste a URL like @MartinDDDD explained in the other forum.

Result: both nodes are GREEN like picture above (so I’m imagining that is “OK” until here

3-Open AI Chat Model Connector
Open AI cht model has a list of Models to choose from.
But I need to pass Gemini Model

So, my intent was to create a String and pass it as a variable.

Open I don’t allow writing Models, only choose from fixed selection

So, I have tried some Models "strings’

• gemini-2.0-flash
• gemini-1.5-flash

**But it ended with the message below **

And I have tried to the next node, and nothing.

Summary: How could you help me? Where am I making mistakes?

[MartinDDDD]

Hey there again,

First of all kudos for the very detailed explanation of what you’ve done so far and what you’ve tried - makes it really easy and enjoyable to help.

I think in general you are on the right way with your set up including passing in a model name via flow variable.

Only thing that might be causing your problem I can see on the screenshots is using LLM Prompter vs. Chat Model Prompter:

Chat Model Prompter

As far as I know LLM Prompter uses completions API structure while Chat Model Prompter uses chat completion structure. If you try and use a model that needs chat completion in LLM Prompter you may get that error.

When making the change be mindful that Chat Model Prompter expects a conversation as in input - so a table with role column (ai or human) and a message column ( with your prompt ) This basic example should help you work that out:

2.1 - Simple ChatGPT prompt – knime

[Felipereis50]

Hello Martin.
I apologize, but I was unable to execute.

I think the first time the OpenAI Authenticator worked was because I hadn’t checked “Verify settings.”

What I mean is that perhaps the initial error is already in the URL or something related to the API.

Am I using the API incorrectly?

1-Do I need to use the User Name or password fields? (Credentials Configuration node)

2-What if?
I pasted the URL wrong? (But I tried, and it didn’t work.)
And should be, something like this below? (but I had tried and did not worked)

Well, I’m lost

API Explanation:

Generating content  |  Gemini API  |  Google AI for...

Getting API

[MartinDDDD]

I see.

I think it is less likely the API key that is the issue.

I looked into the gemini docs and think that:

https://generativelanguage.googleapis.com/v1beta/openai

or

https://generativelanguage.googleapis.com/v1beta/openai/v1

This is based on the code example from here

from openai import OpenAI

client = OpenAI(
    api_key="GEMINI_API_KEY",
    base_url="https://generativelanguage.googleapis.com/v1beta/openai/"
)

response = client.chat.completions.create(
    model="gemini-2.5-flash",
    messages=[
        {"role": "system", "content": "You are a helpful assistant."},
        {
            "role": "user",
            "content": "Explain to me how AI works"
        }
    ]
)

print(response.choices[0].message)

[Felipereis50]

Hello Martin,
Sorry for the delay. This has been a very busy week for me.

Here we go:
I used the code you provided, “https://generativelanguage.googleapis.com/v1beta/openai”, with the “Verify settings” flag, and it went through.
It seems to have worked.

But now I need to configure the model.

As I had mentioned, I passed a string variable with the name “gemini-2.5-flash” or “gemini-1.5-flash”, and I couldn’t get it to work.

I tried putting the string in double quotes, and still nothing.
I’m not able to get it to work.

I would like your help.

Could I possibly give you my API for you to try it on your end?
I can send it via a private message (e-mail) and then delete it after your attempt, what do you think?

If you agree, could you provide me with your email?

By the way, I am using the free Google Gemini Student account. I don’t think that’s a problem.

[MartinDDDD]

Happy to help :).

Before you set me up with an API key try one more thing:

In chat model selection where you pass the flow variable switch the model selection trigger at the top from „default models“ to „all models“?

I think default models triggers some sort of validation against a list of models and if the model in your flow variable is not on the list, it errors out.

Will send you a PM in case you want me to have a crack

[Felipereis50]

BINGOOOOOOOOOOOOOOOOOOOOOOOO

When I use “ALL Models”, various models appear in the list.
And has the “model/” in the string

Great my friend.

Thank youuuuuuuu

t

[MartinDDDD]

Awesome - as always appreciate a solution flag so the next person stumbling across this problem finds the answer quickly!

[MartinDDDD]

BTW - I see the next problem coming: you likely need to use chat model prompter and not LLM Prompter with these models.

[Felipereis50]

For further analysis:

I have configured my Google Cloud, because de API needs a credit cart for billing.
But google give to me a $300 for testing.

This was the Token usage.

I don’t intend to use it much, because I’m still learning.
But it’s a relief that I managed to do it (with your Help)

Now I’m going to look at other examples of flows.

Thanks, Martin