-
Notifications
You must be signed in to change notification settings - Fork 2.3k
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
Fix VertexAI bugs and add function calling #4016
base: main
Are you sure you want to change the base?
Conversation
✅ Deploy Preview for continuedev ready!
To edit notification comments on pull requests, go to your Netlify site configuration. |
Can you add a test in llm.test.ts for the Vertex class? Shouldn't take much, just about 5 lines of code |
Hi are you looking for a TestLLM call? Or something else? |
Hey @sestinj - sorry for the ping - just trying to figure out what kind of tests you would like to see. |
@sestinj - I am looking at llm.test.ts and to add the LLMTest you need a active gcp account. And then it gets a bit tricky. The vertex Ai extension auths via gcloud which is a cli object on your machine. You can't just set a API_KEY like you can for all of the other ones, it would be more complicated. You'd likely have to create a service account and store the credentials in a github secret. Is that what you are looking for when you ask for tests? |
Description
Opening this in place of #3711 as it fixes all of the relevant problems instead of just one.
Seems like a bug was introduced when the embeddings were merged into the llm.
b14b8c4#diff-f6056e365b00006d0872276e5078b8e10b940dfeca7f0aa106b1c842e5aa5a14
The embedding class's base url didn't have a slash at the end, while the llm one did. Neither one was 'wrong', but they were different, so when it was merged, it ended up having a // in the url.
Removing the / before the embedding url fixes this problem.
As well, a bug was introduced in 3b25f78 that broke sending messages as it took the options out of the needed subsection.
Gemini via AI studio and VertexAI are very very similar. So I extracted out the logic that could be copied from the vertexai send and I now call it in VertexAI. This should help prevent any future issues and allow vertexAI's gemini deployment to piggyback on gemini's.
fixes #4166
fixes #3907
fixes #3595
fixes #3303
Checklist
Screenshots
[ For visual changes, include screenshots. ]
Testing instructions
[ For new or modified features, provide step-by-step testing instructions to validate the intended behavior of the change, including any relevant tests to run. ]