-
Notifications
You must be signed in to change notification settings - Fork 7
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
improvements for transferred issues #274
Changes from all commits
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
Original file line number | Diff line number | Diff line change |
---|---|---|
|
@@ -5,6 +5,7 @@ | |
String assignAreaPrompt({ | ||
required String title, | ||
required String body, | ||
String? lastComment, | ||
}) { | ||
return ''' | ||
You are a software engineer on the Dart team at Google. You are responsible for | ||
|
@@ -36,6 +37,7 @@ area-web: Use area-web for Dart web related issues, including the DDC and dart2j | |
Don't make up a new area. | ||
Don't use more than one area- label. | ||
If it's not clear which area the issue should go in, don't apply an area- label. | ||
Take your time when considering which area to triage the issue into. | ||
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I would put even more pressure on the LLM ;) There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. I do wonder if this phrase will make it spend more consideration on the response. I have heard that phrases like 'show your work' will make it produce more tokens, which effectively means that it's using more computation, which may then yield a better result. I'm not sure that it'll help us here since the classification response is only ever going to yield ~1/2 dozen tokens ( |
||
|
||
If the issue is clearly a feature request, then also apply the label 'type-enhancement'. | ||
If the issue is clearly a bug report, then also apply the label 'type-bug'. | ||
|
@@ -48,7 +50,10 @@ Issue follows: | |
|
||
$title | ||
|
||
$body'''; | ||
$body | ||
|
||
${lastComment ?? ''}''' | ||
.trim(); | ||
} | ||
|
||
String summarizeIssuePrompt({ | ||
|
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
(Just curious) Why the limit, if we have a huge context window available?
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
I think this is me being defensive? I can get some data, but I suspect that 90%+ of comments fit in this limit, and the ones that don't are almost always something like a little bit of content and a huge amount of pasted stack trace.
We could keep this as is, raise it substantially so that it really is only protecting against sending in too many tokens, or only have this size limit apply when we're tuning the model (which does have an overall size limit for the tuning data).