1.3.0 · GPL-3.0

Install KnowNote

Two artifacts, both produced by the release workflow from the tagged source. The builds are not code-signed, so the first launch asks you to confirm once on each platform. The steps are below rather than in a footnote.

Open GitHub Releases

Get the app

artifacts

Download from Releases, then follow the one-time step for your platform.

Windowsx64 installer
knownote-1.3.0-setup.exe
SmartScreen will warn once — “More info” → “Run anyway”.
Download for Windows
macOSApple Silicon (arm64)
knownote-1.3.0-arm64.dmg
Unsigned: clear the quarantine flag once after dragging to Applications.
```
sudo xattr -rd com.apple.quarantine /Applications/KnowNote.app
```
Download for macOS

Clearing the first-launch prompt

Neither warning means the download is broken. Both are what an unsigned build looks like.

macOS: “damaged and can’t be opened”Open the .dmg, drag KnowNote into Applications, then clear the quarantine flag once and launch it normally. Only run this on an app you took from this repository’s Releases page.
Windows: “Windows protected your PC”Choose More info, then Run anyway.

First run

setup

Four steps, in the order the app asks for them.

1 · Open it: one promptThe build is unsigned, so macOS quarantines it and Windows shows a SmartScreen warning. Both are one-time, and both are explained below.
2 · Add a model connection: Settings → ModelsNothing is configured out of the box. Add any endpoint that speaks OpenAI, Anthropic or Google, or point it at a local server such as Ollama.
3 · Import something: pdf · docx · pptx · urlCreate a notebook and drop in the material for one question. Parsing and embedding happen on your machine; the embedder downloads once.
4 · Ask: and read the citationAnswers arrive with the passages they were built from, each one carrying where it sat in the document.

What is not shipped

Written down because the alternative is that you find out after downloading.

No Linux build — the release workflow targets macOS and Windows only.
No Intel Mac build — the macOS artefact is arm64.
Nothing is code-signed or notarised.
Audio transcription, quiz generation and slide generation are in development, not in this release.
Early-stage software, maintained by one person, with no support commitment.

The application is early-stage and maintained by one person. The packaging configuration carries a Linux target and the macOS build is arm64 because the release pipeline is run by a single maintainer without an Apple Developer certificate or a Linux machine in the matrix. If you need either, building from source is the path, and an issue saying so is genuinely useful, because it is the only signal that decides what gets built next.

Requirements

what it needs

Windows: x64, installer (.exe)Built by the release workflow on windows-latest.
macOS: Apple Silicon, disk image (.dmg)Built on macos-latest, arm64 only.
A model: any compatible endpointRemote (your own key) or local. Retrieval works without one; answers do not.
Disk: the app, a model cache, your libraryThe embedder is small by design (e5-small, quantised), so a library does not cost gigabytes of vectors.
Network: needed twiceTo download the app, and to download the embedder on first use. After that, retrieval is offline; answers need your endpoint.