ModelReply class sealed
What the model said, in reply to one call: not the transport it
arrived in, nor the pieces streamed beforehand (ReplyArriving,
never recorded) — a driver whose wire is a stream of events gives the
one marking the call as done (such as a result event's own text).
Recorded once, as an outcome.
Either FinishedReply (the model was done) or CutOffReply (the call
ran and spent tokens, but the reply ended before the model was —
such as ChatGPT's own incomplete status). Still recorded as Ran,
never Refused: tokens were spent, and a reply, however partial,
came back. The two are a closed set on purpose: a cut-off reply reads
a caller handles differently from a whole one (it isn't the agent's
full word, isn't what the model actually said in full to post or to
hand back as history), and a type that forces every reader to say
which it means can't be mistaken for the other just because nobody
thought to check a flag.
- Implementers
Properties
- hashCode → int
-
The hash code for this object.
no setterinherited
- runtimeType → Type
-
A representation of the runtime type of the object.
no setterinherited
- text → String
-
The model's text: whole if FinishedReply, only as far as the call
got if CutOffReply.
final
- usage → ModelUsage?
-
What the call spent, when the driver can report it. Null for a
driver that can't (ChatGPT's plan has no cache-creation count or
context window), and for every reply recorded before usage
existed.
final
Methods
-
noSuchMethod(
Invocation invocation) → dynamic -
Invoked when a nonexistent method or property is accessed.
inherited
-
toString(
) → String -
A string representation of this object.
inherited
Operators
-
operator ==(
Object other) → bool -
The equality operator.
inherited