ai-form-fill.

Completed

to

Turn a paragraph of text into values for a web form.

The library reads a form’s fields and select options, builds a JSON schema, and asks a language model to turn the user’s text into matching values. Developers can add it to their own forms and choose the model provider. The thesis release preserves the submitted version, also published as npm 1.0.1.

Work after submission

Later work improves framework integration with native input setters and browser events. The screenshots show an unreleased 2.0 interface with a controller, dictation, review, and undo. Its development source and documentation are separate from the thesis release.

The integration suite runs form-filling scenarios against a real language model. This checks the combined behavior of the library and model, including whether the chosen values match the input text. It does not establish accuracy on every form or replace checks for individual frameworks.

My contribution and AI use

I planned the library and used AI to help resolve isolated implementation issues, closely checking the results. I continued improving it after submission. The linked thesis release preserves the version I submitted, separate from that follow-up work.