In this article we will look at the various configuration settings that administrators can set to control how the AI is used and the information it can return.
*See all Atlas AI & IKS articles here*
For a general overview of the Atlas AI Assistant please see the following article: Atlas AI Assistant
The AI Assistant is the front-end interface for the Atlas Intelligent Knowledge Studio
In this article:
- AI Glossary
- Access the configuration settings
- Control who can use the AI Assistant
- Manage the Knowledge Collection (Atlas 5.2, 5.3 and 5.4 only)
AI Glossary
First we want to supply some definitions as these can be new concepts to a lot of users:
- Generative AI: A type of artificial intelligence that can create new content, such as text, images, or music, based on learning from large datasets of existing material.
- LLM (Large Language Model): An advanced computer program designed to understand and generate human-like text by analyzing vast amounts of written language.
- Prompt: A question or statement given to an AI or computer program, which it uses as a basis to generate a response or perform a task.
- Prompt Engineering: The art and practice of carefully crafting prompts to elicit the most effective and accurate responses from AI models.
- System Prompt or Meta Prompt: A higher-level instruction or query given to an AI system that often guides or structures the way the AI interprets and responds to subsequent user prompts.
- Grounding: The process of linking the AI's language understanding and responses to real-world knowledge, context, or specific data to ensure relevance and accuracy.
-
Fine-Tuning: The process of making small adjustments to an AI model, especially a pre-trained one, to specialize its performance for a specific task or type of data.
Access the configuration settings
To access the configuration settings you will need to be part of either the Atlas ConneX Enterprise Administrators or Atlas IKS Administrators groups.
If you are a member of one of those groups, you can access the AI Assistant settings:
Atlas versions 5.3, 5.3, 5.4 and 6.0
Opening the My Atlas panel from the right-hand side of the main menu, opening Settings and clicking on AI Assistant:
You will be presented with 2 options to configure for the assistant - the System prompt and the Request limits per user:
- We recommend caution when changing the System Prompt since changes may cause the AI to respond in unexpected ways, but if you are confident with prompt engineering then this is where you can change it.
- The token [Facts] found in the prompt refers to the Knowledge Collection(s) available for the AI for your organisation, so by default it can only answer questions about the provided information.
- The limit per user can be very important to control, as you are paying for the required processing in Azure to serve the user requests. This value is the number of requests per user per month. You will need to determine what is the best value here for your organisation based on how much users need to use the assistant.
Atlas version 6.1
Opening the My Atlas panel from the right-hand side of the main menu, opening Settings:
...then 'Grouping' by AI Assistant. The individual settings within the AI Group will appear first as the list displays alphabetically by group, but it is often easier to filter to look at just what you need for this scenario.
There are more settings here in Atlas 6.1. Some of this are not alterable, some are alterable. We advise you understand each setting to comprehend the architecture and functionality behind the settings, but only change the values and 'settings' for the ones highlighted in bold.
- Atlas AI auditing workspace
-
- Url of the AI auditing workspace where sync logs and feedback reports are stored.
-
- Open AI API Version
-
- Version of the Open AI API used by Atlas.
-
-
Include summaries when indexing AI content
-
- If true, during the indexing process a summary will be generated for each chunk and indexed.
-
-
Maximum number of libraries per Knowledge Collection
-
- Determines the maximum number of libraries that can be selected for a single knowledge collection in IKS.
-
- IKS Knowledge Collections Index Name
-
- Azure Search index that contains all knowledge collection names for IKS.
-
- IKS AI Content Index Name
-
- Azure Search index that contains all indexed knowledge for IKS.
-
-
System prompt
-
- An initial input given to the model to guide the generation of text that is always included in each request.
-
-
Non strict grounding system prompt
-
- Part of the system prompt used when a Knowledge Collection is configured to use non strict grounding.
-
-
Post user query system prompt
-
- This is added after the user query on every single message the user sends to the AI Assistant.
-
-
Strict grounding system prompt
-
- Part of the system prompt used when a Knowledge Collection is configured to use strict grounding.
- Part of the system prompt used when a Knowledge Collection is configured to use strict grounding.
-
Control who can use the AI Assistant
Administrators can control who can use the assistant by managing the group Atlas AI Users, so this tool is permissioned. This group can be found in your Azure Active Directory (Microsoft Entra ID).
However, a User will also need access to a Knowledge Collection in order to utilise the AI functionality.
Please view this article for details on AI and IKS permissions and how they interact with each other. This is an important step in your understanding of how to permissions and govern the AI in your correct and required way.
Manage the Knowledge Collection (Atlas 5.2, 5.3 and 5.4 only)
At this time, the assistant only works with a single Knowledge Collection. This is found under the Atlas Configuration workspace in the Global Knowledge Collection library. Due to this restriction you will currently need to copy or move documents or text files containing information into this library. In future releases we will allow multiple Knowledge Collections which can point to your existing workspaces.
Review Sync Logs
In the Atlas Configuration workspace you can access the Knowledge Collection Sync Logs list from the Site Contents. This will allow you to see the syncs (crawls) that have occurred, when they occurred, and who requested them, as well as any errors encountered during the sync. This includes both automatic syncs and user-requested syncs.
Perform a Full Sync of the Knowledge Collection
Please note that the below will change in Atlas 5.2 and crawls will be managed from the AI Assistant section of the My Atlas menu.
If the information you have added to your Knowledge Collection is not being used in responses and you can see that a sync has occurred since you made the change or addition, you may need to perform a Full Sync. This is not the same index as used by SharePoint Search, so we cannot simply reindex the library. Instead, from the Atlas Configuration workspace access the Global Knowledge Collection list you can click the ellipsis (...) in the library bar and choose Sync Knowledge Collection:
After clicking this you will see a confirmation - you need to check the box, then click Confirm to start the full sync:
Please note that performing syncs will incur some Azure processing costs, so this should not be done unless there is a problem with the existing data.
Setting up Terms of Use for Atlas AI
Each user, before being able to use and interact with the Atlas AI Assistant and the content stored within IKS, will need to Accept your Terms and Conditions. These appear over the AI Assistant when first opened and the accepting is stored in the logs for auditing and review purposes.
The Terms of Use are stored in the Atlas Configuration site Atlas Configuration -> Atlas Assets (Document Library) -> AI Assistant (Folder).
You will need to update the context within this document. We advise not to move, delete or replace this document - it's the wording within that is used. Each time this document is updated, it will force each user to re-accept the new terms, so compliance has to be agreed within the 'live' terms.
Comments
0 comments
Please sign in to leave a comment.