Dialogflow CX Data Store Agent - 403 Forbidden on Cited Document

[thesammiller]

Thank you to everyone for all the helpful advice here. I have been able to get a Dialogflow CX Data Store Agent working. However, when users click the cited document, they get a 403 Forbidden. The documents are in Cloud Storage buckets and are not publicly available. 

The users have access to Cloud Storage in IAM, and I even added a custom Cloud Storage read scope to the OAuth. It seems like one answer might be using webhooks to create signed URLs? Are there any other ways to make private documents available to users in Dialogflow CX? 

[xavidop]

Hi @thesammiller , unfortunately not, you will need to create a different public bucket and link it to a new data store. The information in that bucket could be easily accessible and it going to be more safe since you will know which information is public or not.

Another thing that you can try is to have a webhook that transforms those gs bucket links to something else publicly available.

Best,

Xavi

[piyush_garg]

@thesammiller signedURL is one of the good option as their you can define the time at how much time it should be public 

Regards

Piyush Garg

[kevin_PE]

I am running into the same issue. Since you only get the uri, snippet and title, how can you make a signedURL? You don't get the document id, document name or bucket info, any of which would actually make it much easier than trying to strip that info out from the uri. I thought that there would be some way to get the data from the $request.knowledge object, but that doesn't seem to be the case.

[lijokjohn]

Hi @thesammiller ,

Did you find a solution for opening the cited documents? COuld you share some insights?

Thanks

John

[kevin_PE]

It basically was what I had mentioned, manually splitting the stringValue and creating the bucket/file from that:

    // Create presigned URL

    const url = responseData.actionLink.stringValue || "";

    const bucketName = url.split("/")[3];

    const fileName = url.split(bucketName)[1].split("#page=")[0].slice(1);

    const pageNumber = url.split("#page=")[1];

    const signedUrlConfig = {

      version: "v4" as "v4" | "v2" | undefined, // Can also use 'v2' based on desired format

      action: "read" as "read" | "write" | "delete" | "resumable", // Specify action (read, write, or delete)

      expires: Date.now() + 1000 * 60 * 60, // Expires in 1 hour

    };

[lijokjohn]

@kevin_PE 

Thanks. Your solution works.

However later I came across the below article from GCP.

I could get the ' 403 Forbidden' error go away by just switching off the data audit logs. Works only if you are using a authenticated browser though.

https://cloud.google.com/storage/docs/troubleshooting#trouble-download-storage-cloud

Thanks,

Lijo John

Nevertheless I also realized that