Forum Discussion
Copilot accessing a URL to retrieve content - hallucination?
Hi,
When I provide a prompt to Copilot to access a website URL to search for an answer, what technique of RAG does it apply? Computer vision of the rendered web page? Or reading through the source code?
Example prompt:
The goal is to tell me what are the upcoming planned power outages from the Vector website.
The source is: https://help.vector.co.nz/planned
The expectation is you include brief details for each of the planned outages in a list format.
The prompt response I am getting says there are no known planned outages.
1 Reply
When Copilot accesses a URL to retrieve content, it typically reads through the source code of the webpage rather than using computer vision to analyze the rendered page. This allows Copilot to extract text and other relevant information from the HTML content. However, if the information is dynamically generated or requires interaction, Copilot may not be able to access it directly. In such cases, it's important to ensure that the content is accessible and structured in a way that Copilot can interpret.
