-
Notifications
You must be signed in to change notification settings - Fork 360
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
Integrate Vertex AI Models into GuardRails #402
Conversation
…text-bison models
…dling and added user guide
…VertexAI to langchain_google_vertexai.VertexAI due to deprecation
------ | ||
``` | ||
|
||
Now we can test Vertex AI models with GuardRails. The `config` directory includes a variation of the ABC Bot but calling the Gemini 1.0 Pro model from Vertex AI. The main difference is in the `config.yml` file. The relevant section is |
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.
We should not duplicate the config in this folder. The notebook should copy the config when running and update the models part. The config folder itself should also be added to .gitignore
so it doesn't get committed.
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.
Updated in commit 5319dbc5b89201e1e4d7046386183dbd7c3b4a12.
@trebedea : can you please review the changes to the evaluation part? Thanks! |
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 good, just some minor changes and a comment on Langchain usage.
Thanks @aishwaryap ! I've made a few additional changes. Looks good 👍 ! |
Update to PR #387 (which will be closed)
This PR provides support for calling Vertex AI models in GuardRails. Topical evaluation has been done with
gemini-1.0-pro
andtext-bison
models and self check evaluation for hallucination, fact checking and content moderation has been done withgemini-1.0-pro
.Note that GoogleCloudPlatform/generative-ai#344 with Vertex AI models results in calls occasionally erroring out. This PR still contains some very general
try-catch
blocks in hallucination and moderation evaluations to handle this. Errors are reported as a separate category in evaluation.