Learn Anything LLM

Learn Anything LLM is a Tauri 2 desktop app for learning with LLMs. Chat is the home surface, in the ChatGPT / ChatWise idiom: multi-turn tutor runs can retrieve project sources, call web search and specialized tools, and show their activity traces as they go. Everything else — curriculum, practice, mastery, settings — opens from the command palette (⌘K).

Grounded search, two ways

Grounded web search chooses between provider-native search and an application-managed Google-backed retrieval pipeline:

• Native web-search adapters for OpenAI, Anthropic, Gemini, and xAI.
• A Google CSE / explicitly configured SERP proxy fallback.
• One total deadline with cooperative cancellation across routing, retrieval, page fetch, generation, and validation.
• Normalized sources and citations, with structured validation of grounded answers.

Bring your own key

Keys live in a plain TOML file on your machine and are never returned to the frontend — only source and masked hints are exposed. Model capabilities load from a user override, a bundled YAML registry, or a static emergency fallback, intersected with what each adapter certifies; configuration can restrict capabilities but never grant new ones.

Paranoid by default

• Internal page fetches reject local and private destinations and validate every redirect.
• Search-result content is treated as untrusted: grounded generation receives no tools, and evidence URLs are stripped of credential-bearing query parameters.
• The bundled frontend runs under a restrictive Tauri content security policy.

Under the hood

A Rust backend behind Tauri 2 IPC, a React frontend on Vite, Bun for tooling, and a bun run check gate that runs formatting, Rust lint and tests, TypeScript checks, and frontend interaction tests. Local study data and retrieval live on the machine — the cloud is only the model providers you choose.