Knowledge Bases for Amazon Bedrock today supports added accumulation connectors (in preview) – Journal Global Online

Using Knowledge Bases for Amazon Bedrock, groundwork models (FMs) and agents crapper regain contextual aggregation from your company’s clannish accumulation sources for Retrieval Augmented Generation (RAG). RAG helps FMs hit more relevant, accurate, and bespoken responses.

Over the time months, we’ve continuously added choices of embedding models, agent stores, and FMs to Knowledge Bases.

Today, I’m agog to deal that in constituent to Amazon Simple Storage Service (Amazon S3), you crapper today start your scheme domains, Confluence, Salesforce, and SharePoint as accumulation sources to your RAG applications (in preview).

New accumulation maker connectors for scheme domains, Confluence, Salesforce, and SharePoint
By including your scheme domains, you crapper wage your RAG applications admittance to your open data, much as your company’s ethnic media feeds, to compound the relevance, timeliness, and completeness of responses to individual inputs. Using the newborn connectors, you crapper today add your existing consort accumulation sources in Confluence, Salesforce, and SharePoint to your RAG applications.

Let me exhibit you how this works. In the mass examples, I’ll ingest the scheme someone to add a scheme field and start Confluence as a accumulation maker to a noesis base. Connecting Salesforce and SharePoint as accumulation sources follows a kindred pattern.

Add a scheme field as a accumulation source
To wage it a try, manoeuver to the Amazon Bedrock console and create a noesis base. Provide the noesis humble details, including study and description, and create a newborn or ingest an existing assist persona with the germane AWS Identity and Access Management (IAM) permissions.

Then, opt the accumulation maker you poverty to use. I superior Web Crawler.

In the incoming step, I configure the scheme crawler. I start a study and statement for the scheme someone accumulation source. Then, I delimitate the maker URLs. For this demo, I add the come of my AWS News Blog communicator tender that lists every my posts. You crapper add up to decade cum or play saucer URLs of the websites you poverty to crawl.

Optionally, you crapper configure bespoken coding settings and the accumulation redaction contract that defines whether the agent accumulation accumulation module be preserved or deleted when the accumulation maker is deleted. I ready the pick modern settings.

In the sync orbit section, you crapper configure the take of sync domains you poverty to use, the peak sort of URLs to creeping per minute, and lawful countenance patterns to allow or eliminate destined URLs.

After you’re finished with the scheme someone accumulation maker configuration, rank the noesis humble falsehood by selecting an embeddings help and configuring your agent accumulation of choice. You crapper analyse the noesis humble info after creation to guardian the accumulation maker sync status. After the sync is complete, you crapper effort the noesis humble and wager FM responses with scheme URLs as citations.

To create accumulation sources programmatically, you crapper ingest the AWS Command Line Interface (AWS CLI) or AWS SDKs. For cipher examples, analyse discover the Amazon Bedrock User Guide.

Connect Confluence as a accumulation source
Now, let’s superior Confluence as a accumulation maker in the noesis humble setup.

To configure Confluence as a accumulation source, I wage a study and statement for the accumulation maker again, and opt the hosting method, and start the Confluence URL.

To start to Confluence, you crapper opt between humble and OAuth 2.0 authentication. For this demo, I opt Base authentication, which expects a individual study (your Confluence individual statement telecommunicate address) and countersign (Confluence API token). I accumulation the germane credentials in AWS Secrets Manager and opt the secret.

Note: Make trusty that the info study starts with “AmazonBedrock-” and your IAM assist persona for Knowledge Bases has permissions to admittance this info in Secrets Manager.

In the metadata settings, you crapper curb the orbit of noesis you poverty to creeping using lawful countenance allow and eliminate patterns and configure the noesis chunking and parsing strategy.

After you’re finished with the Confluence accumulation maker configuration, rank the noesis humble falsehood by selecting an embeddings help and configuring your agent accumulation of choice.

You crapper analyse the noesis humble info after creation to guardian the accumulation maker sync status. After the sync is complete, you crapper effort the noesis base. For this demo, I hit added whatever fictional gathering notes to my Confluence space. Let’s communicate most the state items from digit of the meetings!

For manual on how to start Salesforce and SharePoint as a accumulation source, analyse discover the Amazon Bedrock User Guide.

Things to know

• Inclusion and banishment filters – All accumulation sources hold body and banishment filters so you crapper hit granular curb over what accumulation is crawled from a presented source.
• Web Crawler – Remember that you staleness exclusive ingest the scheme someone on your possess scheme pages or scheme pages that you hit dominance to crawl.

Now available
The newborn accumulation maker connectors are acquirable today in every AWS Regions where Knowledge Bases for Amazon Bedrock is available. Check the Region list for info and forthcoming updates. To see more most Knowledge Bases, meet the Amazon Bedrock creation page. For pricing details, analyse the Amazon Bedrock pricing page.

Give the newborn accumulation maker connectors a essay in the Amazon Bedrock console today, beam feedback to AWS re:Post for Amazon Bedrock or finished your customary AWS contacts, and vow with the originative AI stuff accord at community.aws.

— Antje