Skip to main content
The Document Panel is the collapsible right-side panel in the chat view. It manages all documents available to the current session.

Uploading a document

1

Open the Document Panel

Click the Paperclip icon in the message input toolbar, or press ⌘D. The Document Panel slides open on the right side of the chat window.
2

Upload

Either:
  • Drag and drop one or more files directly onto the Document Panel, or
  • Click the Upload button and use the file browser to select files
Multiple files can be uploaded at once.
3

Wait for parsing

Each document card shows a Parsing spinner while the background worker extracts text. When parsing completes, the card shows Done and displays the chunk count.Parsing is non-blocking — you can continue chatting while it runs.
4

Confirm the document is active

Newly uploaded documents are automatically toggled active in the current session. The toggle switch on each document card shows its active/inactive state.

Per-session toggle

Each document has a toggle that controls whether it’s injected into the context for the current session. You can:
  • Active (toggle on): Document content is included in the next query’s context
  • Inactive (toggle off): Document is uploaded and parsed, but not injected
Toggling a document does not delete it. The document remains in the library and can be re-activated in any session.

Document card information

Each document card displays:
FieldDescription
FilenameOriginal file name
SizeFile size in KB or MB
StatusParsing / Done / Error
Chunk countNumber of text chunks the document was split into
Active toggleWhether the document is included in the current session

Removing a document

Click the button on a document card to remove it. Removing a document:
  • Deletes its parsed chunks from mira.db
  • Removes it from all sessions it was active in
  • Does not delete the original file from your disk
A confirmation dialog is shown for documents that are active in multiple sessions.

Context usage indicator

The status bar at the bottom shows the total context usage including all active documents. If documents push the context over the engine’s budget, older conversation history is compacted automatically (NAE) or the most relevant chunks are selected (RLM).

Supported Formats

Every format MIRA can parse

Document Chunking

How MIRA splits documents into context-sized pieces