Files and images
Use @ references, drag local paths into the input box, or paste a screenshot to give Mogick the material for a task.
Add material to a message
| Input | How to use it |
|---|---|
| Project file or folder | Type @ at the start of a line or after whitespace, then select an item |
| Local file, folder, or image | Drag it into the input box, or paste an existing local path; recognized paths become reference labels |
| Screenshot or copied image | Copy the image, focus the input box, and press Ctrl+V; clipboard support depends on the platform and terminal |
Check the labels before sending and add a clear request, for example: “Explain the error in this screenshot and compare it with the referenced source file.”
The input already contains an image label, while @ opens file candidates. Select and confirm a file, then add your request before sending. Files shown as candidates have not yet been added to the message.
A label identifies the referenced material; it does not mean every file in a folder has been read. The path must exist and be readable. A multi-line list of path text is not guaranteed to become separate labels. If pasting remains plain text, try @ or paste one existing path.
Image understanding
Two routes are available:
- A model that supports image input can receive images directly. For a custom model, declare
input_modals: ["text", "image"]only when the service actually supports images. - An official llm-store session can use
image.understandwhen the service exposes the image-understanding capability, including when the conversation model is text-only. Available image tools also includeimage.ocrfor text extraction; availability depends on the host service.
Sign in with mogick setup, select an appropriate official model in /models, and ask a concrete question about the image. Signing in alone does not enable official image tools for a custom OpenAI-compatible session. Check /tools if the capability is missing.
For image.understand, the local input limit is 10 MiB and BMP is not supported. Convert to a supported format such as PNG or JPEG if needed. These are tool-specific limits, not universal limits for every multimodal provider. Reading an image through a model sends its content to the configured service.
Read and open results
Replies can include links to local files, folders, and images. Opening behavior depends on the terminal; if a link cannot be activated, use the displayed path in your file manager or editor. Check whether the referenced file still exists.
Long replies show a scrollbar when content overflows. Use the mouse wheel, click/drag the scrollbar, or focus the chat and use PgUp / PgDn.
See TUI basics, Authentication and models, and Built-in tools.