The context is being lost from one prompt to the other and is a mess – I can't translate a manual process I run on LLMs into Gumloop

Get Help
, , ,
1

Help me make this work without losing on my requests.

https://www.gumloop.com/pipeline?workbook_id=7WY71Zewov5tj786cEzznP

  1. First Google Sheets Reader reads the “Latest post to convert” column

  2. Split Text node breaks the text into lines, then Join List Items recombines it (this helps normalize the text format)

  3. The content then flows through a series of Ask AI nodes with Error Shield protection, each writing to different columns:

    • First AI: Refines and cleans the text for clarity
    • Second AI: Restructures into LinkedIn post format
    • Third AI: Generates LinkedIn post hooks based on the restructured post
    • Fourth AI: Creates Twitter-friendly versions

  4. Each result is written back to specific columns using separate Google Sheets Writer nodes:

    • “Clean version”
    • “LinkedIn Post”
    • “LinkedIn Hooks”
    • “Nicolás Twitter posts”

The key difference from my previous explanation is that each transformation has its own dedicated Google Sheets Writer node, and the text goes through a normalization process with Split/Join before processing.

2

Hey @nicolas! If you’re reporting an issue with a flow or an error in a run, please include the run link and make sure it’s shareable so we can take a look.

  1. Find your run link on the history page. Format: https://www.gumloop.com/pipeline?run_id={your_run_id}&workbook_id={workbook_id}

  2. Make it shareable by clicking “Share” → ‘Anyone with the link can view’ in the top-left corner of the flow screen.
    GIF guide

  3. Provide details about the issue—more context helps us troubleshoot faster.

You can find your run history here: https://www.gumloop.com/history

3

Hey @nicolas - I understand you use case but could you share the issue you’re running into? The run link from the history page would be helpful too – instructions above ^

Looking at your workbook one issue that I can see is that you’re passing the same content into both the prompt and context which would likely not yield any valuable AI response.

The prompt is the instructions for the AI (ie. the task you want it to perform) and the context is an optional input to provide the AI more details to help perform the said task.

4

Two main issues:

  1. I can’t seem to make prompts after prompt one work. The prompts are meant to improve HOW the ideas are communicated without changing WHAT the ideas are. Anything past the first prompt creates a whole different thing from version 1 (clean) and version 2 (after the first writing prompt).

  2. I don’t understand the issue you are referring to. I’m passing the same context across every writing prompt, but I’m running different prompts / instructions across every writing prompt.

  3. It’s important to mention that I’m comparing the results that I’m getting not only to the apparent following of instructions but to the results that I get if I manually go to ChatGPT and take version 1 through prompt one, and then take the new version through prompt two, and then take that version through prompt 3, etc.

  4. The workbook link is the following: https://www.gumloop.com/pipeline?workbook_id=7WY71Zewov5tj786cEzznP

  5. I’m aware that I’m not paying for the full version, so maybe the results could improve with, say, o1, HOWEVER, it seems to me that I’m not being able to recreate the manual process that I do with GPT because the results past the first prompt are a mess.

5

  1. Ensure you’re passing the response of the previous AI node in the Context input and specify your requirements in the prompt.

  2. When you connect an input to the Prompt then that input replaces the prompt you’ve set in the node manually. In order to use the prompt set on the node, ensure you’re only passing the AI response in the Context input. Read the doc above to learn more about the difference between the two.

image
image3022×1402 291 KB

  1. To clarify, all AI models are available on the free plan, they just cost different amount of credits. Standard models are 20, normal models are 2 and expert models are 30 credits.