Using Gemini as a chat model to read Gmail but not getting results

Gemini (or any chat model) does not automatically access your Gmail data, even if you have authenticated the Gmail API. Here's what's happening:

1. "Sign in with Google" only authorizes the Gmail node, not the chat model.
2. Gemini needs to be supplied with the actual email content as input – it doesn't retrieve it on its own.
3. The response "I'm sorry, I don't have the capability..." indicates that Gemini was not provided with sufficient context.

You'll need to manually input email content from Gmail into the Gemini model.

1- Utilize the Gmail Node to retrieve unread emails or list messages (e.g., Get Messages).
2- Format the Email Content, converting it into a string.
3- Pass the formatted string as a prompt to Gemini 1.5 Flash, instructing it on the desired action (such as summarizing, extracting dates, or replying).

Thank you for taking the time to review my question and provide a response. I've included some screenshots of my workflow; please let me know if there's a more efficient way to share this information.

CaptureAgent436×655 27.1 KB

CaptureGem451×484 17.2 KB

CaptureGmail454×702 22.7 KB

CaptureMain899×502 32.1 KB

I was following a tutorial video, with the only difference being that I am using Gemini as the model instead of Anthropic.

The video does not cover formatting email content, converting to a string, or similar steps. Am I missing some crucial stages in my workflow?

When testing the workflow, does the Gmail Get All Messages node successfully retrieve your emails?

It says “Workflow executed successfully.”

I've switched the Gemini Chat model for the OpenAI Chat model (as recommended in the video), and it's now functioning correctly. I'm uncertain why one works while the other doesn't, but I suspect Gemini might not possess the same email reading capabilities as OpenAI. Thanks again for your input.

1. Utilize a Gmail node to retrieve email content initially.
   Confirm that you are employing a Gmail > Read Email (or comparable) node prior to the Gemini node, and configure it to fetch the message body intended for analysis.

2. Input the email data into the Gemini prompt.
   Within the Gemini node, refer to the content from the Gmail node in your prompt as follows:
   Here’s an email I received: {{$json["body"]}}
   Could you please provide a summary?

Apologies, but I don't quite grasp your instructions. From what I can see, there isn't a prompt window within the Gemini node in callin.io.

Could you please clarify your request?

Hi, I think I can assist.

Could you please share your workflow? Here's a guide.

Hi Grot, this is the workflow that isn't functioning:

When I substitute the Google Gemini Chat node with the OpenAI Chat node, it functions correctly when I instruct the agent to retrieve information about unread emails, for instance.

Ah, I understand. The issue might stem from Gemini 1.5 being a less capable model. The LLM is responsible for invoking the tool usage.

You can enhance this by adding instructions to your system prompt or user message to always call the tool, which I'd suggest doing.

Additionally, consider switching the model to this one: models/gemini-2.5-flash-preview-04-17. It performs significantly better.

Thanks! Switching to Gemini-2.5 resolved the issue. While it's not performing as well as OpenAI, I can now retrieve emails.

Glad to hear it.

Also, I recommend reviewing this function calling benchmark. Currently, the OpenAI models lead the list.

Regards