Train AI - Build your own fine tune AI model

You can use Google Sheets to create a fine-tuned AI model. Open Google Sheets and add or import the training dataset > click Extensions > click Promptrepo > click Train AI > Promptrepo sidebar widget will be displayed > Enter the AI model name, click Next > select input columns, click Next > select output columns, click Next > Click Publish and let Promptrepo fine tune the AI model.



Supported input types

When you select the input columns, you can specify the data type.


  1. String - Simple text, a sentence or a paragraph

  2. File - Document or image

  3. JSON - Structured data format

Supported output types

When you select the output columns, you can specify the data type and also whether empty values are allowed.


  1. String - Generates a free form text

  2. Number - Outputs a numerical value

  3. Boolean - Returns true or false

  4. Enum - Selects from a predefined set of categories

  5. Array - Returns a list of multiple values

  6. JSON - Outputs structured data in JSON format


Advanced settings

Use generative AI: You can enable the Use generative AI option to automatically generate new outputs by recognizing and applying learned patterns from existing data. For instance, if real nutritional data for a food item is unavailable, the model can estimate its nutritional values based on similar foods, ingredient composition, and established dietary trends.


System message: You can include a prompt to establish clear instructions that define the rules and context for how the model should generate outputs. This ensures consistency, accuracy, and relevance by guiding the model’s behavior, tone, and decision-making process.

Create your own AI model using no-code model builder

If you would like to create an AI model from scratch, you can use the no-code AI model builder. Go to promptrepo.com > click on the Try it free button > select the type of data to extract from (ex: Document, Email, Text) > upload the document > Promptrepo will analyze the document and extract the input and output fields > review the results and click Publish to deploy your model.




Made with formfacade