Would it be possible to have an external folder as source for the context grounding? Think I have a team that provides services for my company, they have all their procedure and stuff in a sharepoint folder. Would I need copy it into an index? (Sure the indexer automation could do this?)
Good question, and yes, you could upload files to the index source (the “storage bucket) using the indexing robot, or through the orchestrator API. Metering logic (the payment model) is not 100% finalized yet (as far as I know), but will undoubtedly be a consumption based model using tokens known as AI Units.
Thank you Jeppe for this content . i have couple of questions. Firstly when Context Grounding isn't able to find a highly confident corresponding answer in the provided data, how does it respond . Next how is my data stored or shared with Context Grounding?
As far storage and sharing, data in indexes is stored inside the UiPath ecosystem, but agreements have been made with 3rd party model providers that data is not stored, used for training and not even logged when coming from the UiPath platform. The first part of your question...... not sure really. This stuff is new to me, too :-)
Good point Parvati, It will respond as per the data you have provided. I think if Context grounding is not able to find answers in the provided data then we should improve the quality of data. Whatever you send for indexing will be used for response. Higher the quality, better the response.
Impressive!
Yeah, it’s not bad, and this is still only in Public Preview. More exciting stuff to come, no doubt. 🙂👍
Awesome Video, Thanks Jeppe
Would it be possible to have an external folder as source for the context grounding? Think I have a team that provides services for my company, they have all their procedure and stuff in a sharepoint folder. Would I need copy it into an index? (Sure the indexer automation could do this?)
Good question, and yes, you could upload files to the index source (the “storage bucket) using the indexing robot, or through the orchestrator API. Metering logic (the payment model) is not 100% finalized yet (as far as I know), but will undoubtedly be a consumption based model using tokens known as AI Units.
Thank you Jeppe for this content . i have couple of questions.
Firstly when Context Grounding isn't able to find a highly confident corresponding answer in the provided data, how does it respond . Next how is my data stored or shared with Context Grounding?
As far storage and sharing, data in indexes is stored inside the UiPath ecosystem, but agreements have been made with 3rd party model providers that data is not stored, used for training and not even logged when coming from the UiPath platform. The first part of your question...... not sure really. This stuff is new to me, too :-)
@@UiPathwithJeppe thanks for the reply.
@@ParvatiAya-su8jo thanks for your support 🙏🙂
Good point Parvati, It will respond as per the data you have provided. I think if Context grounding is not able to find answers in the provided data then we should improve the quality of data. Whatever you send for indexing will be used for response. Higher the quality, better the response.
I'd assume it's checking first the RAG source. If it doesn't find a plausible response, it then returns whatever the model you are using would reply.