The choice of LLM host for the LLM Chatbot custom visual lies entirely with the user. To utilize the chatbot, users must create an account with their preferred LLM host. Currently, options include OpenAI and Azure OpenAI, with Azure OpenAI offering enhanced data protection through robust enterprise-grade security measures, read more here.
To use the chatbot, users must have an account with their chosen LLM host and possess an API key. All interactions with the chatbot are securely processed through the selected service's REST API using the HTTPS protocol. No data is ever shared with other platforms or services, ensuring complete privacy. This approach guarantees robust data protection and secure communication at all times.
To start, the LLM Chatbot visual performs most file-related operations locally. However, to harness the full potential of LLMs, the file content is initially sent—once per session—to the LLM host to generate embeddings (learn more about Embeddings - OpenAI). These embeddings, which transform the text into numerical vectors, are stored in the browser's memory.
From that point onward, semantic searches are conducted locally within the browser. Only the necessary snippets of text are sent to the LLM host to answer questions in natural language or summarize file contents. Because the text content is transmitted to the LLM host through a secure environment, and the document itself is never uploaded online, using the LLM Chatbot visual to analyze documents is as secure as interacting with the chatbot directly.
Interacting with datasets through natural language offers a highly intuitive and powerful way to extract insights. To ensure the security of this feature, the LLM Chatbot visual queries datasets using the Power BI API. These queries are automatically generated as DAX queries based on user input. The results are securely processed in the browser, interacting exclusively with the selected LLM host and no other services.
Since using the Power BI API to execute DAX queries requires authentication, users must first set up an application in the Azure cloud. This application is then used to grant the necessary permissions to run DAX queries. This approach ensures seamless dataset interaction while maintaining strict data privacy and security.