Context is what Nodespell Agent knows about a request before it starts working. You provide context by including selected nodes, mentioning nodes with @, and attaching reference files. Clear context makes Agent faster and more accurate, because it doesn’t have to search the Project to guess which nodes you mean.
What context does Agent get by default?
Section titled “What context does Agent get by default?”Agent does not automatically receive everything you have selected on the canvas. Each message carries only these things:
| Context | How it gets into a message |
|---|---|
| Nodes you include | Include selected nodes, or a node chip you leave in the message |
| Nodes you mention | @ followed by a node name |
| Files you attach | Attach files, drag and drop, or paste |
| The conversation so far | Earlier messages in the same conversation |
| Where you are | The current Level and the size of the graph |
When Agent needs more, it uses its own read-only lookups. Agent can search the current Level or the whole Project, inspect nodes, and follow connections upstream or downstream. Each lookup appears as a step in the conversation, for example Searched the project for “portrait”. For Levels, see Levels and Folders.
A follow-up that points back with words such as “it”, “them”, or “the same”, for example “make it brighter”, reuses the nodes from your previous request. This happens only when the follow-up doesn’t include nodes of its own.
Include selected nodes in a message
Section titled “Include selected nodes in a message”-
Select nodes on the canvas. A chip appears above the message box. With one node selected, it reads Include followed by the node’s name. With several, it reads Include 3 selected nodes.
-
Select the chip, or open + and choose Include selected nodes. Each node becomes a removable chip with a thumbnail.
-
Remove any chip you don’t want by selecting its ×.
-
Type your request and send it.
One message can include up to 60 nodes. If you select more, only the first 60 are included. The tooltip warns you before you add them. After three chips, the remaining nodes collapse into a +N more chip that you can expand.
When you pick one of the suggested requests in a new conversation while nodes are selected, Agent includes the selection for you.
Mention a node with @
Section titled “Mention a node with @”Type @ in the message box, or open + and choose Mention a node. A list of nodes on the current Level appears. Each row shows the node’s name, its type, and a thumbnail. Keep typing to filter the list, which shows up to eight matches. Press Enter or Tab to insert the highlighted node, or press Escape to close the list.
Mentions work well when the node isn’t selected, or when several nodes have similar names.
Attach reference files
Section titled “Attach reference files”Open + and choose Attach files, drag files onto the message box, or paste them. Agent uploads each file to the Project’s Assets in the Uploads folder. Agent then passes the file to the request as an Asset reference. Agent can place that Asset on the graph and connect it. For the Asset library, see Asset Manager.
| Limit | Value |
|---|---|
| Files per message | Up to 12 |
| Size per file | 200 MB or smaller |
| Supported types | The same types you can upload to Assets. See Supported media. |
Each attachment chip shows its status: Queued, upload progress as a percentage, Processing, Ready, or Failed. You can send the message only after every attachment is ready. If an upload fails, select the retry button on the chip or remove the file. If you send attachments without any text, Agent receives Use the attached references. as the request.
Choose a reference role
Section titled “Choose a reference role”Each attachment chip has a role menu that tells Agent how to use the file. New attachments start as Source.
| Role | Use it for |
|---|---|
| Source | The material to work on directly |
| Character | A person or character to keep consistent |
| Style | A look, palette, or rendering style |
| Environment | A location or setting |
| Motion | Movement or camera reference |
| Audio | Sound, voice, or music reference |
| Other | Anything else; describe it in your message |
The role is a hint. Agent decides how to wire the file into the graph, and you can see the result on the canvas.
Frequently asked questions
Section titled “Frequently asked questions”Does selecting nodes on the canvas send them to Agent?
Section titled “Does selecting nodes on the canvas send them to Agent?”No. Selecting nodes shows an Include selected nodes chip, but the nodes reach Agent only when you include them or mention them with @.
How many files can I attach to an Agent message?
Section titled “How many files can I attach to an Agent message?”You can attach up to 12 files at a time, and each file must be 200 MB or smaller.
Where do attached files go?
Section titled “Where do attached files go?”Agent uploads attached files to the Project’s Assets in the Uploads folder. They stay there after the conversation ends, like any other uploaded Asset.
Can Agent find nodes I didn’t mention?
Section titled “Can Agent find nodes I didn’t mention?”Yes. Agent can search the current Level or the whole Project by name and content, and it shows each search as a step. Naming the node yourself is faster and avoids mistakes when names are similar.
Related
Section titled “Related”- Agent overview — open Agent and send a first request.
- Modes and approvals — what Agent may do with that context.
- What Agent can do — request examples that use mentions and attachments.
- Upload and organize Assets — manage the files Agent uploads.
- Supported media — file types you can attach.