项目的 README,来自 github.com/amosblomqvist/learnharness。合并为一个 markdown 文件:README.md。
An agent harness for learning.
Download · How it works · Documentation · Issues · Discord
LearnHarness uses AI agents to plan and teach lessons from your current understanding to your goal. It builds a detailed picture of what you understand and how you reason. That picture is updated as you learn.
Instead of just giving you an explanation, it uses your answers and thinking to decide what you need next. Teaching builds understanding one reasoning step at a time. Practice makes it usable. A persistent knowledge graph carries it across sessions.

Think out loud. The thinking fields let you explain your reasoning, uncertainties, and reactions in your own words. This is how the agent gets a detailed picture rather than just a record of right and wrong answers. You can type or use dictation.
Bring your files. In the desktop app, attach PDFs, photos, or any other file with the paperclip, or drag them in. The file stays where it is: the message carries its path, and the agent reads it from there.
The agent can also research questions in the background, create visual explanations, show your own images in the lesson and mark them up with arrows and labels, and build interactive demonstrations.
What you are taught accumulates as propositions and reasoning chains: individual claims and the premises that lead to them. The graph shows this structure, with links back to the source sessions. You can also import notes.
Future teaching sessions can read the graph instead of starting from scratch. Practice sessions can draw their material from it. It stores the ideas and their connections, not ratings of your ability.

Download from learnharness.ai · GitHub Releases
.deb.Choose your models, then start a session with what you want to understand. You can use provider API keys, supported subscription connections in the desktop app, an installed Claude Code CLI, or a custom OpenAI-compatible endpoint, including a local model server. Optional dictation and web search use separate service keys. Web research can use a Tavily, Brave, Exa, Firecrawl, or Ollama key in Settings → Providers → Search. Ollama search uses its hosted service, not a local model server.
See Getting started for model setup, dictation, files, and notes imports, and context, caching, and costs for advanced cache preferences.
All three platforms support in-app updates.
npm ci
npm run plugins:setup
npm run dev # http://localhost:3000
npm test # fixtures and fake providers
See the documentation index for all guides, development for setup and checks, architecture for the runtime, keyboard shortcuts, themes for the theme file format, and desktop builds and releases.
For trusted local plugins, start with the authoring guide and capability map. It links the question, practice-problem, submission, view-context and UI contracts. Use examples/plugins/ for small runnable templates, plugins/ for feature source projects, and the host architecture for internals.
Questions, bug reports, and feature requests: GitHub Issues.
Security vulnerabilities: report privately as described in SECURITY.md, not in public issues.
The app is free and open forever. If we ever charge for anything, it will be for convenience — managed keys, hosting — never for features you can run locally.