Two options to avoid manual creation of corresponding Pega Knowledge articles:

1) Use the Pega Knowledge bulk import (.csv file) feature to pre-load all of the targeted Sharepoint articles, assuming they each have unique absolute URL's.  In the .CSV file, the ContentID column could be left "blank", which would result in Pega Knowledge creating new content ID's during the import.  The ContentType column should specify "URL", and then each ContentBody cell would contain the Sharepoint URL's for each article (each article representing a row in the .CSV file.

2) A second approach involved more of an integration effort, where either the existing Sharepoint index of the repository is made available to Pega's Elastic search engine, so when an article search or suggestion is being made, Elastic would parse both indexes (Pega Knowledge index and the Sharepoint index) and "collate" the results, determining the most relevant results from both indexes to return to the search or suggested content function in the UI.  Note that Pega does not provide a spidering/crawler capability, so that may be required if the solution requires crawling and indexing the Sharepoint repository.