Scripted Prompting

In LMQL, prompts are not just static text, but can also contain control flow (e.g. loops, conditions, function calls). This facilitates dynamic prompt construction, but also allows LMQL queries to respond dynamically to intermediate model output.

For instance, let’s say we want to generate a packing list. One way to do this would be with the following query:

Open In Playground

sample(temperature=0.8)
   "A few things not to forget when going to the sea (not travelling): \n"
   "[LIST]"
from
   'openai/text-ada-001'

A list of things not to forget when going to the sea (not travelling):

-A phone with call, texting or tech services
-A copy of the local paper
-A pen or phone Keytar

This can work well, however, it is unclear if the model will always produce a well-structured list of things. Further, we have to parse the response to separate it into separate items and process it further.

To address this, we can provide a more rigid prompt template, where we already provide the output format and let the model only fill in the THING variable:

Open In Playground

sample(temperature=0.8)
   "A list of things not to forget when going to the sea (not travelling): \n"
   "-[THING]"
   "-[THING]"
   "-[THING]"
   "-[THING]"
   "-[THING]"
from
   'openai/text-ada-001'
where
   STOPS_AT(THING, "\n")

A list of things not to forget when going to the sea (not travelling):
-A phone with a/r text
-pletter
-accoon
-Films about/of the sea
-A has been in the Poconos for/ Entered the Poconos