Announcements
Hi, we're looking for a clean and easy way to get frequently updated data into our bot to be used for generative answers without needing to upload new files with each update.We have some OneDrive for Business accounts and I'm thinking we can just pass it a directory with word documents and it could do generative answers on the data in thd docs. Is this feasible? If so, how would I implement it? Is it just pasting the link to the directory somewhere? Ideally, we ultimately would want it setup so that the users wouldn't need to authenticate in order to have access to the drive documents we link.I have also been looking at indexing the OneDrive with Azure AI Search and accessing the index through the API to return results. But it appears OneDrive isn't natively indexable without a 3rd party, paid connector. Probably not the right forum for this, but if anyone has any insight it may also be helpful. Also just open to seeing this completely differently.Thanks for any help!
What is the main challenge here? Getting files uploaded automatically to OneDrive? How are those files being generated to begin with?
The main challenge is that we want to have dynamic documents. So the information can get to the bot without a user needing to upload changed documents.
If OneDrive is where you store documents, it should be pretty straightforward automating document ingestion with Power Automate.
Thanks @adilei. I guess the consideration is returning the appropriate information. With Generative Answers it seems like the uploaded document data is stored in dataverse, but also indexed and returns appropriate chunks of data for the answer, similar to how Azure AI Search seems to do. Which makes me think that's what's going on under the hood, that it's using Azure AI Search. That seems to be how it works with Sharepoint and I assume OneDrive for Business. Giving it a directory and it makes each document in all the subdirectories searchable. I've seen this shown for Sharepoint and it was implied it worked the same for OneDrive.But if I just retrieve documents from OneDrive, how would I be sure to return the appropriate text, assuming I'm using Custom Data for Generative Answers. I can reference specific documents, but if they have large amounts of data, the signal to noise ratio would be pretty high and based on previous experience with GPT prompts and lots of tokens in them it'd likely make it less reliable in finding the appropriate responses to the users inquiries.Maybe there's some other way to do this? That's where the idea of running an AI Search Index on the data came from and then querying the REST API to return only relevant data to be used as custom data for generative answers.Thanks for your support on this! ✨
Hi @wswitzer ,
based on your use case (having multiple documents regularly uploaded to your bot and with anonymous access) I think it is best that you directly upload your documents in the bot (in the generative AI menu) - that way it will store those documents in Dataverse and use semantic indexing + Azure cognitive search on them to find the best data to answer any questions.
Contrary to what you stated uploading documents on Sharepoint / OneDrive only makes them available to the graph API, it does not chunk / index them like on Dataverse (unless you are using the new M365 Copilot but that is another discussion), this kind of indexing is not ideal for large/complex documents.
I have tried both (documents on Sharepoint/OneDrive VS Dataverse) and the quality of RAG from Dataverse is far better (which is normal - Sharepoint only leverage the graph API, not a full chunking+indexing)
You could also automatically update/sync those documents in your Copilot using a power automate flow (they are stored in the Dataverse table "Chatbot subcomponents")
Hope his helps -
That is super helpful. Thanks!I noticed in that table a bunch of things called "Bot file attachment", I'm assuming these are the uploaded documents. But there's no name for the document or unique identifier. Are this the right data? And does it just overwrite the previous document with the same name if it gets synced using the given PowerAutomate flow?Thanks for your help!
Hi @wswitzer you are correct, files are stored in the column filedata with name Bot File Attachment in the column ComponentType.
When you upload several times the same file the web app will create a new name and delete the previous attachment (to only keep 1 record with the file attached - that's why only 1 line has the attachment icon in my screenshot). See an example where I uploaded 3 times the same licensing guide:
Not sure if that is something you'd have to do manually when uploading a new document using Power Automate - that's something to test.
Under review
Thank you for your reply! To ensure a great experience for everyone, your content is awaiting approval by our Community Managers. Please check back later.
Congratulations to our 2026 Super Users!
Congratulations to our 2025 community superstars!
These are the community rock stars!
Stay up to date on forum activity by subscribing.
Valantis 601
chiaraalina 137 Super User 2026 Season 1
Haque 133
