When you first enter Atlas, you will see the homepage, once you have created a workspace it will show on this page. To get started, click ‘New Workspace’ in the top right of the screen, give your workspace a name, and click ‘Create’. This opens a new chat where you can begin asking questions.

When you move to a new question, you can start a new workspace so earlier context does not mix into your new request.

To start a new conversation navigate back to the homepage using the ‘Workspaces’ button in the top left of the screen and once you are there click on ‘New Workspace’ just as you did at the beginning.

Each new chat begins with no memory of earlier conversations. The agent will have no knowledge of prior questions, answers, and context from previous chats.

If you need a detail from an earlier chat, you may ask the agent in the earlier chat for a summary, and then either copy and paste the text or upload as a file into the new chat.

Very long chats may lose quality as the agent reaches the limit of its context window. In this case, it may be advised to start a new chat and import the key details.

The homepage is where you can access all your workspaces. From here you can open a previous workspace to continue or view a conversation, as well as rename or delete workspaces to keep your list organised.

To write a good prompt, it helps to understand what Atlas CI is doing to answer your query.

When you send a prompt, Atlas CI breaks your query into a plan for how to search Norstella’s curated enterprise dataset and extract and analyze the information, before generating an output that answers its interpretation of your query.

The clearer and better structured your request, the more easily the agent can work out what you want, where to search for the answer within its connected database, and how to structure its response.

Use these points to help it understand you:

It’s important to remember that the agent works only from the data it is connected to or has been trained on. It does not know your internal context, documents, or anything outside its competitive intelligence database and background knowledge from its LLM training.

Do not assume it has context if you have not given it. Context can be provided to the agent either through the prompt itself or by uploading a file with contextual details into the chat.

An artifact is a self-contained output that the agent produces during a chat, such as a table, chart, or downloadable file. Artifacts are embedded within the agents response, and appear a button with a chart or table icon and the name of the artifact. Examples of artifacts within a chat can be seen in image below highlighted with red boxes.

To open an artifact simply click on it and it will open in a panel to the righthand side of the chat.

You can also view all artifacts that have been created within a workspace using the “View all Artifacts” button in the left hand panel. From here you can click on any artifact and it will open the artifact.

Starring an artifact marks it as a favorite so you can return to it without scrolling back through the conversation. It is most useful in long chats with several questions and follow-ups, where the outputs you want are hard to find later. As you read through a response, star the artifacts you expect to use so you can locate and download them when you need them.

You can star an artifact in two ways:

Artifacts created by Atlas CI are available to download as editable PPT files so you can easily incorporate them into your workflow.

You can download directly from an open artifact by clicking the download icon in the top-right.

You can also download an artifact from the “View All Artifacts window” by clicking on the download icon next to the artifact that you want to download.

For long chats with multiple responses, you may want to create an organised list of the most useful artifacts to download by starring your artifacts.

You can attach a file so the agent can use your data or context alongside Norstella’s data when answering your question. The upload feature supports PDF, DOCX, DOC, XLSX, XLS, CSV, TXT, HTML, JPEG, JPG, PNG, MD, GIF and WEBP file types up to 100MB per file.

Different ways to upload:

1. When you start a new chat. In the message bar that reads "Ask Atlas CI", select "+ Attach Files", then choose "Upload file". An "Add Files" window opens where you can drag and drop a file or select "Browse" to pick one. Select "Done". The file then appears as a labelled chip above the message bar (for example "Test.xlsx"), with a "Files Attached" count shown in the bar. Type your question and send.

In a follow-up question. The "+ Attach Files" button stays available in the message bar throughout the conversation, so you can attach a file to any follow-up the same way. Use this when a later question needs a different or additional file.

2. From the left-hand panel. The "Uploaded" section lists files you have already uploaded and includes an upload icon for adding new ones.

3. Once you have uploaded a file you will see a "View All Uploads" button in the “Uploaded” section which opens the "All Uploads" window, where you can search and manage every file you have uploaded. You can also upload additional files from here.

In the "View All Uploads" pop-up, each file has actions next to it, including an "Ask in Chat" icon. Selecting "Ask in Chat" automatically attaches that file to the message bar, so you can type a question and submit a query against it without re-uploading.

To remove a file from a prompt, select the × on its chip above the message bar before sending.

To delete an uploaded file entirely, use the delete (trash) icon in the "All Uploads" window.