AI Suite is the instance-wide store for AI resources: written instructions, reference files, tool definitions, MCP server registrations, and the memory the model maintains for itself. One set of resources serves every AI processor on the instance, so you upload a file once and then select it in each processor that needs it.

Uploading a file adds it to a section. It affects nothing until you select it in a processor.

To access the AI Suite page:

Navigate to Analytics > AI Suite.

The left panel groups the resource types by the job they do. Each entry shows its file count.

Content that shapes what information the model has for your AI Analytics flows. Instructions, Skills, and Context all accept files, and they differ in what you use them for. Put standing rules that apply to every prompt in Instructions. Put a procedure for one kind of job in Skills. Put reference material the model reads in Context.

Things the model can call during a prompt.

Agent files are stored, listed, previewed, and included in an export, but no processor uses them yet. Agent runtime behavior arrives in a later release.

Each section offers two ways to add a file. Select Upload to add one from your computer, or New file to write one directly on the page. Select a file to open its details panel and preview its contents.

A processor that answers questions about pump maintenance needs three files:

You upload each file to its section once, then select all three in the AI Responses processor. A second processor on the same instance can select the manual alone.

Litmus Edge (LE) stores these files in a single folder tree, with a subfolder per resource type. Every resource is identified by its file path, so renaming a file produces a new path rather than moving an existing one.

When you configure an AI Responses processor, each resource type appears as a list of checkboxes populated from these sections. Selecting a file stores its path in the processor configuration, and the processor reads the selected paths each time it runs.

Every section is optional. A processor with nothing selected sends its own prompt and the conversation history, and nothing else.

Two consequences follow from paths being the identifier:

LE validates files when you save them, and what it checks depends on the section.

A registration names the server, its URL, the headers to send, an optional list of allowed tools, a timeout, and whether to accept insecure TLS. LE connects over Streamable HTTP.

After a registration saves, LE tests the connection and shows the result on the page. A successful test lists the tools the server currently offers, each with the description it reports. If the server cannot be reached, the page shows a warning naming the error. The registration is still saved.

Select Test connection to run the test again at any time.

Read the tool descriptions. They come from the remote server and go into the model's context, so they shape what the model does.

Use Allowed tools in a registration to expose only some of a server's tools. Without it, every tool the server offers is available to any processor that selects the file.

Export packages the whole resource tree as a ZIP file. Use it to back up a configuration or to move one to another instance.

Import merges a suite ZIP into the existing tree. Files that already exist are overwritten, and the import reports which entries it wrote and which it skipped, with a reason for each.