# Annapurna

Source: https://docs.rubrik.com/en-us/saas/ai/annapurna.html

---

# Annapurna

Rubrik Security Cloud includes the Annapurna application which you can use to obtain answers from your data, using generative AI.

Annapurna is an enterprise application. By using the generative AI capabilities of Annapurna, you can leverage your protected data to obtain natural language answers to business questions through retrieval-augmented generation (RAG).

You can use Annapurna in two different ways: 

  * Custom chatbots can be created in Rubrik Security Cloud (RSC) to process your prompts through the large language model (LLM) of Azure OpenAI Service.
  * Data chunks generated via RAG can be retrieved using an API endpoint, and these chunks can then be sent to your specified LLM within your custom GenAI application.

To begin using Annapurna with your enterprise data, your administrators create chatbots. Each chatbot specifies which workload data to index and which sensitive data filters to apply. A chatbot can be configured to exclude sensitive user data based on your Data Security Posture policies. Data sources from the following workloads can be included:

  * Microsoft 365
  * Jira Cloud
  * Salesforce
  * Confluence[1](#fntarg_1)

After specifying the data to include, your administrators assign a role to the chatbot. Only users with the assigned role will be able to use the chatbot. The data available to a user through a chatbot depends on the access permissions the user has for that data. Annapurna uses the permissions assigned by each source application to determine the access a user has to data from that application.

Once chatbots are created, your enterprise users can launch a chatbot and ask natural language questions about the data available through the chatbot. Based on the query, Annapurna uses RAG to get relevant data from the available data sources. The retrieved data and the query are combined into a prompt and sent to the LLM. The user gets a natural language response to their query that is based on the data included in the prompt. This process does not involve training the LLM.

The chatbots that your administrators create can use multiple data sources from different workload types. This means a chatbot can query data from various workloads at the same time. Queries through such a chatbot can give answers that combine data from different sources.

Important:

Annapurna uses generative AI to respond to your prompts. AI-generated content may contain errors or inaccuracies. Independent verification of such content is strongly recommended.

[1](#fnsrc_1) To enable Confluence as a data source, contact Rubrik Support.