miniExtensions’ ChatGPT automation lets you use OpenAI models to analyze, rewrite, or generate content from Airtable records and save each response back to your base. Give the automation a prompt, optionally include documents or images from attachment fields, and choose where the result should be stored.
What you can do
- Summarize, rewrite, translate, or classify record content.
- Use shared instructions or a prompt built from each record.
- Analyze supported documents and images from selected attachment fields.
- Save the response to your chosen output field.
Prompts and attachments
A prompt can stay the same for every record or come from an Airtable field. A formula is useful when you want to combine instructions with several record values, such as a product name, description, and desired tone. System Instructions define the overall task and response style.
You can also select attachment fields so a compatible model can analyze their documents or images. For example, ask it to summarize a report or identify details in a supplied document. The request uses the information you supply, rather than automatically reading your entire base.
The response is saved in one output field. If you ask for several pieces of information, they are returned together rather than automatically filling separate columns. Use the available skip or overwrite settings and an Airtable view to control which records are processed.
Examples of useful workflows
Turn feedback into a consistent summary
Combine a feedback field with a request for a short summary and a category. Store the model’s response in an output field for review. If you need separate Airtable columns, plan how the returned text will be parsed; asking for JSON alone does not create or populate several fields automatically.
Extract information from documents
Select an attachment field and ask the model to identify the details you need, such as a document’s subject or key dates. Review the response against the source file before using it downstream.
Draft content from record data
Build a prompt field containing a product name, description, and desired tone. The model can draft copy into a separate field while the original data remains available for comparison.
How to get started
Connect your base, add your OpenAI API key, and choose your prompt, optional attachments, model, and output field. Test a record before processing more.
Good to know
Your OpenAI API key and provider usage charges apply. The model uses the prompt and attachments you supply; review AI-generated results before relying on them.
