The OneSearch AI Assistant was developed as a tool to aid researchers in developing more effective searches of the SDSU Library OneSearch database.
Once a research question is entered to the form a system prompt running on an Open WebUI API endpoint using the Llama 3.1:70b model converts this question into a search string that includes the Boolean operators AND, OR, NOT, and common database search operators, like quotation marks, asterisks, and parentheses.
This search string is then used to query the OneSearch API to retrieve the number of results and the first five journal article citations and abstracts found.
A second system prompt then uses the article abstracts to create a summary of the type of information being found.
The search string can then be modified by the researcher and re-run in the application, or can be used to directly query the library OneSearch database for books, articles, and other information types.
This application was created by Lucy Campbell, Electronic & Continuing Resources Librarian, and Keven Jeffery, Digital Technologies Librarian, from the Library Content Organization and Management Unit.