GPT Unleashed - Free ChatGPT Addon for Google Sheets

Community
, , , ,
1

Dear community,

As a heavy user of Google Sheets and ChatGPT, I have tried many add-ons, but those that I tested all lacked some features and aspects I felt would provide more value: Most were paid, offered little to no control over the models (and model parameters) used, and I really wished I could manage and reuse prompts with good outputs as templates, e.g. for prompt engineering. In a nutshell: I wished to have more flexibility & options for prompt organization.

Therefore, I set a goal to improve my coding skills and integrate all these features into a free add-on.

:checkered_flag: Two days ago, my add-on was finally published in the Google Workspace Marketplace.

You can get it here:

(Edit: thx for giving me permission to post this :heartpulse:)

Since I have only a few users so far, I would greatly appreciate any feedback to improve the add-on.

Functionality:

  • Sheets Formula:
    =GPTUL(**input** [or single prompt as text input], **job** [template shortcode or text input], **context** [template shortcode or text input], **apiQueryPreset** [optional preset shortcode, otherwise default settings are used])

  • Job Template (example):
    In the settings, you could set the prompt for the shortcode “extract-nouns” as: “You’re a noun extractor. You extract all nouns from the user’s input text and return a comma-separated list, without further comments.” Now you can reuse this shortcode for any input, e.g.: =GPTUL(A2, "**extract-nouns**"). If your input for the job parameter does not match a shortcode, the input will be interpreted as an independent prompt. This way, you can test various “Job prompts” before saving them as a shortcode: =GPTUL(A2, "Extract all verbs from the input text and return without further comments.").

  • Context Templates
    … work the same way. You can combine them with Job Prompts or leave them out. A context can greatly improve your output by providing GPT with information about your project or intentions. For example, a context prompt could be: “Your outputs are being used for data entry workflows. Return accurate results without further comments, as they are used programmatically.”

  • apiQueryPreset: You can change model parameters like the temperature or max_tokens in the settings and save them with a shortcode. You can then use this shortcode in your formula. Let’s say you want to A/B test the outputs of GPT-4 with a temperature of 0.9 vs 0.2. In this case, you could save one preset as “4-02” and one as “4-09”. Then, you’d be able to compare outputs of formulas by using these in your formula as parameters: =GPTUL(A2, "extract-nouns",, "**4-02**") vs. =GPTUL(A2, "extract-nouns",, "**4-09**").

Thanks a lot for your time and feedback :slight_smile:

1 Like