AI Summarization
Version 8.0+
You can use Yarkon to Summarize Documents using AI. Any text or PDF document that a user has access to can be summarized using one of the AI model available through AWS Bedrock.
Yarkon is using AWS Bedrock for AI Summarization. Therefore, only subscribers who use AWS S3 can use this feature. In other words, AI Summarization is not supported when using S3 compatible storage providers.
Models
Yarkon supports the following foundational models:
- Titan Text G1 - Express by Amazon
- Nova Lite by Amazon
- Llama 3 by Meta
These models were chosen because they provide a sensible compromise between the quality of the summary, the speed of processing users' requests, and overall usage costs. In the future, additional models will be added, after we validate their performance.
If you are using Yarkon Server, you can add other models. Please contact us for details on how to do so.
Set Up
Enable Models in AWS Bedrock
Start with making sure your organization has access to at least one of the models listed. Follow this document from Amazon for the required steps.
Grant IAM Permissions
To grant Yarkon users access to AI Summarization, you'd need the following IAM policy statement:
{
"Statement": [
{
"Effect": "Allow",
"Action": "bedrock:InvokeModel",
"Resource": "*"
}
]
}
You can make this policy more specific by changing the
Resource
to only allow specific models.
Add this IAM policy statement to the main Yarkon policy. You can either create a new IAM policy and attach it, or make the changes inline in the Yarkon policy. If you are using the Integrated Security Model, you'd need to add the policy (again, attached or inlined) to each user/group/role policy used by an end-user you'd like to allow AI summarizations.
Verify that it Works
Test that it works using the Yarkon Admin Console:
Finally, log in as a user, and verify that the feature works as expected: