3.2.4 Upload files and screenshots
The AI Dock accepts files, images, and screenshots as reference. Whether images are analyzed directly depends on whether the current model supports vision.
Supported attachment types
| Type | Supported | Use |
|---|---|---|
| Images | png, jpg, jpeg, webp, gif | Show Moss screenshots, UI, or error dialogs |
| Documents | txt, py, yaml, json, sh, xml, md, cpp, h, log, etc. | Give Moss file contents |
| Screenshots | Paste or upload as an image | Quickly show what you see |
Three ways to add attachments
| Method | Action |
|---|---|
| Attachment button | Paperclip next to the AI Dock input opens the file picker |
| Drag and drop | Drag from the file manager onto the input |
| Paste screenshot | Paste clipboard image into the input |
Attachments appear above the input and are sent with the message.
Typical scenarios
Analyze an error screenshot
When you see a GUI error on the board, paste a screenshot into the AI Dock and ask, for example: “What does this error mean and how do I fix it?” Vision-capable models read the image; Moss combines it with device context.
Upload a config for edits
Upload launch files, YAML, etc., and describe the change you want. Moss reads the content, proposes edits, and writes back after you confirm.
Long logs: summarize and locate
Save logs as text, upload, and ask: “Find the errors and summarize in time order.” Moss pulls the key events.
Notes
Attachments become part of the model input and consume tokens:
- Larger images and longer text cost more.
- Resending the same file spends tokens again.
- Prefer uploading once, then referring to it in follow-ups.
If the token pill near the top of the AI Dock nears the limit, start a new session or trim attachments.