Fill a screen's inputs from my words
Fill a screen's inputs from the words a person typed, such as timer 25 min or lunch with Ruth tomorrow at 1, by each input's format, choices or type, with no model
Fill a screen's inputs from the words a person typed, such as timer 25 min or lunch with Ruth tomorrow at 1, by each input's format, choices or type, with no model
Give schema (the screen's input schema, { properties }), words, and now (the device's time, ISO) for words like tomorrow or in 20 minutes; used ([[start, end]]) are the spans of the words that picked the screen, which are not values; choices names labels for enum values. Answers values (only what the words give and the schema accepts; anything uncertain is left for the person) with how each was read. Parsers and their words are src/input-parsers.json.
Nothing: your AI starts it with your words.
Connect the AI you already use to Trillion once. Then ask in your own words; your AI finds this tool by what it does and runs it.
Say to your AI: “Use Trillion to fill a screen's inputs from the words a person typed, such as timer 25 min or lunch with Ruth tomorrow at 1, by each input's format, choices or type, with no model”