-
Notifications
You must be signed in to change notification settings - Fork 2.3k
Feat: Add prompt injection detection settings UI + update logging #4651
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
Feat: Add prompt injection detection settings UI + update logging #4651
Conversation
6c7866a to
82a0c31
Compare
8cd5566 to
82a0c31
Compare
8393cf5 to
6562601
Compare
DOsinga
left a comment
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 had a quick look, but I fear this feels rather vibe codey - there's some duplication, unrelated changes, I wonder if you could do a bit of a review more yourself?
You're right, I wanted to get it done too quickly. Having another look now and will do some cleanup. Thanks for the comments |
4bd65ae to
1c89c0a
Compare
1c89c0a to
a79ec4e
Compare
56bad4b to
f57511e
Compare
9b5c660 to
7a9d9d4
Compare
bd7293a to
0ae76d5
Compare
ui/desktop/src/components/settings/security/SecuritySection.tsx
Outdated
Show resolved
Hide resolved
|
looks like linting issues now |
|
so the config is now saved as: not in a security: section - is that expected? (also I noted it didnt' save the threshold until I changed it, which I assume means default?) @dorien-koelemeijer there are a few other changes not GUI related, what is best way to check this is still healthy? looks ok otherwise, @DOsinga happier with shape of code? |
|
thread safety/oncelock stuff looks good too |
It'll be security.threshold: xx and security.enabled: true/false. But I think most people will just go through the UI settings now, since it's a lot easier. Re: non-GUI related changes, you can simply test the changes in goose CLI/desktop version locally. I've tested until the last lint commit and everything was still fine - will do some final testing in the next couple of hours, but it should all be good. |
|
@dorien-koelemeijer but when I saw it in the config.yaml - it wasn't grouped under security (was an underscore) - just wanted to know if that was expected? |
Really? Let me have a look as well 👀 In any case, I'll have to update instructions once this is merged. Would be great to get both this PR and the other one together in one release to prevent confusion |
I think I tried to keep things backwards compatible first and still had that config saved - you're right, it's definitely with underscores
Seems like it would make sense to keep underscores, since all config is saved like that? I guess it depends on whether this PR gets released at the same time as the other one, then we don't have to have things be backwards compatible. Honestly, probably easiest to keep it in line with all other config though and use underscores? |
Have tested both desktop and CLI version - all seems fine still. Testing will be the same as for the other PR if you wanted to give it a go as well to be safe |
DOsinga
left a comment
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.
Looks much better, yeah! I think you can delete even more LLM comments - I always do
* main: docs: Change community page sections (block#4984) docs: remove temporary Hacktoberfest issue templates (block#4982) Create multi-channel researcher prompt (block#4947) docs: Add Community Content section to Community Page (block#4964) Allow empty API Key when registering custom provider (block#4977) Feat: Add prompt injection detection settings UI + update logging (block#4651) Make create_session work concurrently (block#4954) Lifei/create save recipe to file (block#4895)
…ock#4651) Signed-off-by: Itz-Agasta <[email protected]>

PR Description
UI configuration for prompt injection detection
Allow users to enable prompt injection detection and the confidence threshold through the UI instead of having to manually update goose config file.
Update security logging
Some general cleanup
Remove some unnecessary/unused code