Question objects
questions maps an id you choose to a question object. Every object has a type and an instructions string.
noul
A statement; the answer is P(true).
| Field | Type | Required | Description |
|---|---|---|---|
type |
"noul" |
yes | |
instructions |
string | yes | A statement, e.g. "The task is finished" |
criteria |
object | no | {"true": "...", "false": "..."}: what each side means |
JSON
{"type": "noul", "instructions": "The next action is risky",
"criteria": {"true": "it deletes data or moves money", "false": "it only reads"}}choice
One of several options.
| Field | Type | Required | Description |
|---|---|---|---|
type |
"choice" |
yes | |
instructions |
string | yes | The question |
criteria |
object | yes | {option_key: description}, 2 to 256 options |
args |
object | no | {option_key: {arg_name: fill or extract question}}: arguments generated for the chosen option only |
JSON
{"type": "choice", "instructions": "Which team should handle this ticket?",
"criteria": {"billing": "refunds and payments", "shipping": "delivery and tracking", "tech": "website and app problems"}}score
A level on an ordered scale.
| Field | Type | Required | Description |
|---|---|---|---|
type |
"score" |
yes | |
instructions |
string | yes | The question |
criteria |
array | yes | Levels, lowest first |
JSON
{"type": "score", "instructions": "How upset is the customer?", "criteria": ["calm", "annoyed", "angry", "furious"]}fill
Write a short value, constrained to its type. Preview.
| Field | Type | Required | Default | Description |
|---|---|---|---|---|
type |
"fill" |
yes | ||
instructions |
string | yes | What to write | |
kind |
string | no | "text" |
text, int, number or json_list |
pattern |
string | no | A regex the value must match. Only with kind: "text" |
|
max_tokens |
integer | no | 16 | 1 to 256 |
JSON
{"type": "fill", "instructions": "the amount due, in dollars", "kind": "number", "max_tokens": 16}extract
Copy a value from the state, verbatim where possible. Preview.
| Field | Type | Required | Default | Description |
|---|---|---|---|---|
type |
"extract" |
yes | ||
instructions |
string | yes | What to extract | |
kind |
string | no | "text" |
text (or coordinates, see locate) |
many |
boolean | no | false |
Return every match as a list |
list_format |
string | no | "json": read a many extract from a JSON list instead of one item per line |
|
max_tokens |
integer | no | 64 | 1 to 512 |
JSON
{"type": "extract", "instructions": "every order number mentioned", "kind": "text", "many": true, "max_tokens": 64}locate
Where something is in an image. A locate is an extract with kind: "coordinates". Preview.
| Field | Type | Required | Default | Description |
|---|---|---|---|---|
type |
"extract" |
yes | ||
kind |
"coordinates" |
yes | ||
instructions |
string | yes | What to find | |
image |
integer | no | 1 | Which image, 1-based (the N of <image:N>) |
scale |
string | no | "rel1000" |
How coordinates are expressed to the model: rel1000 or pixels. Answers are in pixels either way |
max_tokens |
integer | no | 64 |
JSON
{"type": "extract", "kind": "coordinates", "instructions": "the Pay button", "image": 1, "scale": "rel1000", "max_tokens": 64}