LibreMuse

LibreMuse is a LibreOffice Writer extension that brings AI-assisted text editing directly into your sidebar. Connects to any OpenAI-compatible API — Ollama, LM Studio, or cloud LLMs. Provide context from selected text or the full document, then rephrase, summarize, or proofread.

Quickstart

Download the latest extension file (.oxt) from the releases page. Open LibreOffice Writer, go to Tools > Extension Manager > Add, and select the downloaded .oxt file to install.

Open the sidebar in LibreOffice; you should see the LibreMuse icon for the extension. Click the icon to open the extension panel.

How it works

The sidebar has two panels — Main and Settings.

Main panel:

• Enter a prompt (e.g. "Fix grammar", "Summarise this", "Rewrite in a formal tone") and click Generate.
• If text is selected in the document, the AI response replaces the selection.
• If nothing is selected, the response is inserted at the cursor position.

Providing context:

• If text is selected and no other context is set, the selected text automatically serves as context for the LLM (in addition to being replaced by the response).
• Click Copy selected to explicitly copy the current text selection into the context area (useful when you want context from one place but are editing another).
• Or check Use document as context to send the entire document as reference material to the LLM.
• Click Clear to remove the stored context.

Context is passed to the model as reference material — the AI uses it to understand the topic but does not repeat or paraphrase it in the output.

Settings

Open the Settings panel to configure the extension.

• Model ID — the model name (e.g. gpt-4o, mistral/mistral-medium, gemma3:1b)
• Model URL — the endpoint URL (e.g. https://api.openai.com/v1/chat/completions, http://localhost:11434/api/chat)
• Model API key — your API key (optional for local servers that don't require auth)
• Custom system prompt — override the default system prompt that guides the AI's behaviour
• Test — verify connectivity to the configured endpoint. The button automatically tries common URL paths (/v1/chat/completions, /api/chat) if the entered URL doesn't respond directly.
• Save all — persists settings to LibreMuseConfig.json in the LibreOffice user config directory.

The extension sends requests directly to the configured endpoint — no intermediary backend.

Ollama

Set the Model ID to the Ollama model name (e.g. gemma3:1b), and the Model URL to http://localhost:11434/api/chat (or leave it empty — the Test button will auto-detect it).

LM Studio and other OpenAI-compatible servers

Any OpenAI-compatible chat endpoint works (LM Studio, llama.cpp, vLLM, etc). Set the Model URL to the server's chat completions endpoint.

For LM Studio, that URL is http://127.0.0.1:1234/v1/chat/completions (not its native /api/v1/chat).

Build

Requires ../unodit/unodit.py (sibling directory).

make package

Produces dist/LibreMuse.oxt.

No automated tests, linter, or CI. Verify code manually.

License

See license.txt for license.