ChatGPT sometimes won't give a result with no error

[Miroslav_Horvath]

Hello, I am having an issue with the ChatGPT module.

I have a scenario where I receive a message from ChatGPT, which then goes to another ChatGPT module that processes the message and subsequently sends it to email.

The problem is that the second ChatGPT module has a ±60% chance of not sending (creating) a “result” even though it receives the input correctly. The images show both outputs without any changes to the scenario.

image 1 with no result

image 2 with result → correct response

Two things are interesting (atleast in my eyes). The first is that it does not cause any error. The second is that despite no changes in the scenario, sometimes the run is successful and sometimes it is not.

I would like to ask if you know why this is happening, or how to create a loop that would repeat the second module (without having to repeat the entire scenario, because I would burn through a large number of tokens).

The problem arises when a reasoning model is involved, regardless of which one. Does anyone know what to do about this?

Thanks.

[samliew]

Welcome to the Make community!

We’d love to help, however, you have not provided sufficient information to demonstrate the problem that would allow us to reproduce the scenario and any issue/s and/or error/s.

To allow others to assist you with your scenario, please provide the following:

1. All Relevant Screenshots

We need to see what you’re working with to give you the best advice. Screenshots are extremely important because Make is a visual editor — a picture provides us with more context.

It would help us identify the issue by having screenshots of:

• the full scenario,
• the zoomed in section of the scenario that you are referring to (if your scenario is huge),
• a full scenario run/execution (showing output “speech” bubbles),
• any error messages,
• individual module fields (inputs/values),
• relevant filter settings (conditions), and
• each module’s output bundles
• any external services (spreadsheet headers, sample data, regex101.com, etc.)

You can upload images here using the Upload icon in the text editor:

We would appreciate it if you could upload screenshots here instead of linking to them outside of the forum. This allows us to zoom in on the image when clicked, and avoid tracking cookies from third-party websites.

2. Scenario Blueprint

Please export the scenario blueprint. Providing your scenario blueprint file will allow others to quickly recreate and see how you have set up the mappings in each module, and also allows us take screenshots or provide module exports of any solutions we have for you in return - this would greatly benefit you in implementing our suggestions as you can simply paste module exports back into your scenario editor!

To export your scenario blueprint, click the three dots at the bottom of the editor then choose ‘Export Blueprint’.

You can upload the file here by clicking on this button:

3. Module Output Bundles

Please provide the output bundles of each of the relevant modules by running the scenario (you can also get this without re-running your scenario from the History tab).

Click on the white speech bubbles on the top-right of each module and select “Download input/output bundles”.

A. Upload as a Text File

Save each bundle contents in a plain text editor (without formatting) as a bundle.txt file.

You can upload the file here by clicking on this button:

B. Insert as Formatted Code Block

If you are unable to upload files on this forum, alternatively you can paste the formatted bundles.
Here are some ways to provide text content in a way that it won’t be modified by the forum.

• Method 1: Type code fence manually —
  Add three backticks ``` in a separate line before and after the content, like this,

  ```
  text goes here
  ```
  

• Method 2: Highlight and click the “preformatted text” button in the editor —
  

• Method 3: Upload your file and share the public link —
  (this method is only recommended for large files exceeding the forum upload limit)

Providing the input/output bundles will allow others to replicate what is going on in the scenario, especially if there are complex data structures (nested arrays and collections) or if external services are involved, and help you with mapping the raw property names from collections.

Sharing these details will make it easier for others to assist you.

[Miroslav_Horvath]

Hello,

I found the answer. When I was preparing all the information I wanted to send here, the second model did not give me an answer 10 times in a row. It was a reasoning model with a high reasoning value set.

The output token limit was 5000 so that the resulting report from the module would not be too long.
But then I noticed a magic sentence in the module: “When using reasoning models such as the o1 models, this value is the sum of reasoning + output tokens.”
With a limit of 5000, the reasoning was long and exhausted the limit without a response. Once I removed the limit, everything worked as it should.

But otherwise, thank you for your reply.