Kendra Salesforce connector only indexing document titles?

0

I am playing around with using Kendra as the knowledge store for a RAG approach to genAI as recommended here: https://aws.amazon.com/blogs/machine-learning/quickly-build-high-accuracy-generative-ai-applications-on-enterprise-data-using-amazon-kendra-langchain-and-large-language-models/

One of the datasources I want to use are Knowledge documents from our Salesforce instance, so I used the Kendra Salesforce Connector v2 to index our knowledge documents.

Unfortunately, although the indexing was successful it appears that the only information indexed from the documents were the document titles. (I have drawn this conclusion because the only excerpts I have been able to retrieve via the retrieval api have been the titles)

This seems almost useless. Am I missing something?

pez
已提问 10 个月前327 查看次数
1 回答
0

Hi,

Did you add all required permissions to access all data and manage articles in SFDC docs:

Added credentials to your Salesforce server for a user with read-only access
 to Salesforce by cloning the ReadOnly profile and then adding the View All Data 
and Manage Articles permissions. These credentials identify the user making the 
connection and the Salesforce connected app that Amazon Kendra connects to.

See https://docs.aws.amazon.com/kendra/latest/dg/data-source-v2-salesforce.html#supported-features-v2-salesforce

Best,

Didier

profile pictureAWS
专家
已回答 10 个月前
  • Hi Didier,

    Thanks for your response!

    For this experiment I just gave the connector my own SF credentials. I do have both the mentioned permissions.

    Are you confirming that I should expect to see more than the document titles as excerpts?

    I am seeing a link to the full document in Kendra, but I need to be able to retrieve relevant excerpts from the document body from Kendra.

您未登录。 登录 发布回答。

一个好的回答可以清楚地解答问题和提供建设性反馈,并能促进提问者的职业发展。

回答问题的准则