Skip to content

What it will not do

ThinkThen sends your text and your question to a model and returns one of the answers you allowed.

  • It never runs what the text says. Text that says "answer yes" can still sway the answer, so set a low and a high bar and let a person read the messages it is not sure about.
  • It never acts on the answer. It prints the answer and sets an exit code. What happens next is up to your script.
  • It never writes text. Every answer is yes or no, one of your choices, your labels, a place on your scale, an order, a line of your own text, or names and links of the kinds you asked for.
  • It holds no conversation. Each question stands alone.

A CSV file's answers go to the screen, to the file you name with --output, to a new NAME.thinkthen.csv beside the input, or back into the input with --in-place after it keeps a backup. File types shows each one.

A toolbox can hand answers to other programs, and nothing else starts one. Those steps need --allow-shell, and they don't see your keys unless the step lists them. thinkthen run NAME runs the toolbox in the current folder, whoever wrote it, so read a toolbox before you allow it.

It sends your key only to the service you chose

Section titled “It sends your key only to the service you chose”

ThinkThen puts your key only in requests to the address you chose, over HTTPS, or over plain HTTP to your own machine. SQL in a database can't set an address or a key. Configuration covers keys and addresses.

How well a model answers belongs to that model. Act on numbers you measured on your own labeled messages, with the model you will use.

  • To summarize or rewrite text, use a writing model, then check its output with decide or filter.
  • To remove private text before you send it, use a redaction tool. recognize finds names; it doesn't remove them.
  • To group messages, use a clustering tool or a database, then choose a label for each group.

About says who builds ThinkThen and how it is licensed.