Add backend controller and client for Gen AI #57944
Conversation
| return render(status: :ok, json: {status: filter_result.type, flagged_content: filter_result.content}) if filter_result | ||
| inputs = params[:inputs] | ||
| temperature = params[:temperature] | ||
| # TODO: Format input to send to Sagemaker. |
There was a problem hiding this comment.
Still waiting clarification on how to format. Using this ref as a reference for now.
There was a problem hiding this comment.
Sorry, kind of got comment happy :) Mostly small things / me trying to understand this stuff better!
One overall q -- is there any value in maintaining the existing API that can actually talk to an AI (ie, openai) in the short term, rather than cutting over to this new version that isn't actually hooked up? I was wondering if Dan and co. might find that useful for taking screenshots/etc.
Thanks Ben - this is all great feedback! Posted your overall question in this Slack thread and the response from Dan and Sanchit is that we should go ahead and proceed with getting |
This PR adds the
aichat_controllerwhich will make requests to Sagemaker when the API is available and the corresponding client fileaichatCompletionApiwhich makes a call to the controller. This will be used when a user submits a chat message in the Aichat workspace.The controller receives:
savedAiCustomizationschatContextthat includes the newest user message text,userId,currentLevelId,scriptId, andchannelId.The controller sends back a dummy assistant response for now. Screenshot from dev console after adding print statements in
aichatCompletionApi.tsandaichatReduxofresponse.Response displayed in workspace:
Links
jira
Testing story
Tested locally using /allthethings levels.
Screenshots from Network Tab:
Deployment strategy
Follow-up work
Privacy
Security
Caching
PR Checklist: