There are no any news extracted in email

andyw · August 21, 2025, 1:43am

Describe the problem/error/question

Attached is the workflow that it used to

monitor the RSS sites and see any AI related news
extracted related AI related news and send email

What is the error message (if any)?

Actually, there are no any error messages when running. The email also be received.

But all content is about no AI related news. (See attached email pictures)

Any setting wrong?

Please share your workflow

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:

jabbson · August 21, 2025, 2:06am

Hey @andyw hope all is good.

Try to change the gemini chat model on the bottom workflow to use gemini-2.5-flash-lite model. It seems to be able to make tool calls:

I also changed the prompt a bit. See if this works better for you.

andyw · August 21, 2025, 3:34am

Hi @jabbson

Thanks for your help. I change the model to

Gemini-2.5-pro-preview-06-25

But I get the message

Error in sub-node ‘Google Gemini Chat Model‘

[GoogleGenerativeAI Error]: Error fetching from https://generativelanguage.googleapis.com/v1beta/models/gemini-2.5-pro-preview-06-05:streamGenerateContent?alt=sse: [429 Too Many Requests] Gemini 2.5 Pro Preview doesn’t have a free quota tier. For more information on this error, head to: Rate limits | Gemini API | Google AI for Developers.

Is it because too many posts for AI agent to handle?

If so, how to limit the post for processes, e.g. only 2 latest posts for each RSS sources?

thanks

Regards

AW

jabbson · August 21, 2025, 3:57am

The reason for the rate limit is clearly indicated in the error message. Not sure if your key is attached to the billed project, this could be the reason, since it mentions something about not having a free tier. The main point - not all models can work (or work well) with tools. It sometimes takes some research and trial and error, to find the one which does. Also tried with openai, worked just fine too.

jabbson · August 21, 2025, 4:31am

When retrieving the news from the vector db, the node is doing similarity search, and then returns N most relevant resources, which are passed to the LLM for a single call to compile the answer.

Unless you are asking about the ingestion of article and not retrieving. If this is the case, you can loop over rss urls and limit the rss feed retrieval like so:

This will process all rss feeds but only yield 12 articles (2 per each of 6 urls).

jabbson · August 21, 2025, 2:08pm

You can literally ask the agent to generate the title for you as for the link, you need to be ingesting it in the fist step.

great question, you could install it as npm package if you are self-hosting and use from the Code node.

jabbson · August 22, 2025, 2:17am

What do you mean by “new”? why do you need the new one? Did you break the old one?

find one here:

github.com/jreyesr/n8n-nodes-docxtemplater

samples/Docxtemplater_sample.json

master

{
  "name": "Docxtemplater sample",
  "nodes": [
    {
      "parameters": {},
      "type": "n8n-nodes-base.manualTrigger",
      "typeVersion": 1,
      "position": [
        0,
        0
      ],
      "id": "db1c73e1-2772-4e6a-b419-89b6b27a40ae",
      "name": "When clicking ‘Test workflow’"
    },
    {
      "parameters": {
        "method": "POST",
        "url": "https://demo.gotenberg.dev/forms/libreoffice/convert",
        "sendBody": true,
        "contentType": "multipart-form-data",

This file has been truncated. show original

andyw · August 22, 2025, 2:49am

Thanks @jabbson

I not delete “Normalize Field 1” and there are url and title in the node. that why i believe the information already be ready in db.

When run the “News reader AI“ node, there are only news summary but not see the title and url.

My question is how to include all other information like url and title in the email send out?

The system message from “News reader AI” node

Only get last week’s news. Act as a tech news aggregator and write in plain, easy-to-understand English. Prioritize news related to the following topics: {{ $json.Interests }}.
If none of those topics are mentioned in the news, use your best judgment to highlight the most newsworthy, frequently mentioned and relevant events in technology.

Provide a total of {{ $json[‘Number of news items to include’] }} news items.

Thanks

Regards

AW

jabbson · August 22, 2025, 3:10am

Try asking for it explicitly:

Only get last week’s news. Act as a tech news aggregator and write in plain, easy-to-understand English. Prioritize news related to the following topics: {{ $json.Interests }}. Always include the article URL in the answer.

If none of those topics are mentioned in the news, use your best judgment to highlight the most newsworthy, frequently mentioned and relevant events in technology.
Provide a total of {{ $json[‘Number of news items to include’] }} news items.

andyw · August 22, 2025, 9:25am

I’m a beginner with n8n and not a technical user, so I may ask some basic questions as I learn.

The purpose of this workflow is to:

Extract AI-related news using an AI Agent.
Send a newsletter email (this part is working fine).
Generate a PowerPoint file containing two AI news items, where each item includes a title, url, and summary, placed in predefined locations in the slide.
→ I’m using Docxtemplater for this part.

Questions

1. Markdown to JSON Node Issue
The “News Reader AI” successfully generates the news content. However, the “Markdown to JSON” node fails to extract the url and summary fields.
→ Any suggestions on how to modify the code or regex to correctly extract these fields?

2. Docxtemplater Node Design
I’m using a Merge (Combine) node to merge the PowerPoint template binary file with the news JSON data before passing it to the Docxtemplater node.
→ Is this the correct design approach for generating a PPT with dynamic content?

system · November 20, 2025, 9:57am

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.