Library centralizes reusable conversation assets — images, video, audio, PDF and Word documents, or external links. Upload and organize them, then the agent decides on its own which one (if any) fits a given moment in the conversation.
How the agent picks and sends what you uploaded
The agent doesn't get a fixed keyword mapping — it matches an asset's name and description against the conversation using plain word matching (exact word, shared prefix, or common stem), so write those two fields like you're describing the resource to a person, not tagging it for a search engine.
- Up to 2 attachments per AI turn — if the model tries to reference more, only the first two go out.
- Each file type has a real size ceiling for direct delivery through Meta: 8 MB for images, 25 MB for video, audio and documents. Go over that and Setterix automatically falls back to sending a shareable link instead of the file itself, so oversized assets still reach the lead — just differently than you might expect.
- If the model attaches a file and also typed its bare filename as a message bubble, that redundant bubble is dropped automatically so the lead doesn't see the filename as a separate text line.
Good practices
- Name resources by their real use case, and write the description the way you'd explain it to someone — that text is literally what drives matching.
- Keep links, prices and product files current.
- Remove obsolete versions.
- Test delivery in the simulator before relying on it in a live flow.