Hi I'm experimenting with the AI Agent Studio object and setup a test case to read excel file, but when it tries to use the Semantic Search Tool I always receive this response.An error occurred while executing the tool [SemanticSearch]: The user [username] does not have sufficient privileges to perform the requested action.And return error from the tool. Is there anything that need to be setup or allowed?
Discussion posts and replies are publicly visible
Are you passing the document to the AI Agent as input or is it configured as a Knowledge Tool for the agent? Also, does your prompt contains any instruction to verify if the User is valid or not - if so make sure to pass the User record in knowledge tool.
Passing the excel file as Input from the object to test the agent. And no there is no instructions to check if the user is valid.
Can you check the security of underlying tools. The initiator or process executing the agent must have access to involved tools in the agent else we get permission issues. So verify security of agent, tools within agent as well as user/process and share how it goes.
Hi Harsha I'm doing that in the designer menu where i just created the object and trying to use the AI Agent studio Test function where you can execute test scenario this is where I'm getting that. As i mention the user which is me has all of the access of the application. There is no Semantic Search Tool to find anywhere in Appian i think this is the default studio tool for reading documents and I don't see a way to adjust access to it.
I meant the Tools(process models) we enable Agent to take action with. I guess you dont have that, and the semantic search tool, being inherent to AI Agent and not configurable) is not something we can fix.
Overall the error suggests security/permissions missing so event though you are a system administrator with all the permissions, suggest you to configure 'security' (administrator and view group) for AI Agent and involved process models/document folders.
If its all set properly already then try to set a simple action prompt for agent with text/integer inputs instead of document to see if agent executes properly.