Generates hands-on, multi-part technical tutorials on demand using LLMs, targeting developers and researchers who prefer learning by doing. It addresses the scarcity of tutorials in niche technical areas by providing an interactive, local learning experience that fosters deep understanding, recreating the "aha!" moments often missed when LLMs automate tasks.

How It Works

The project integrates LLM "skills" within coding agents for tutorial generation, extension, and verification, managed by a Go CLI. Tutorials are served via a local web UI, positioning LLMs as tutors rather than task automators. Transparency is key: tutorials document their sources, LLM model, and writing "voice," promoting verifiable learning.

Quick Start & Requirements

Installation is via a single binary: macOS (brew install devenjarvis/tap/lathe), Linux (curl -sSf https://raw.githubusercontent.com/devenjarvis/lathe/main/install.sh | sh), or Go/source build (go install github.com/devenjarvis/lathe@latest). Requires Go 1.25+ for Go installs. Crucially, Lathe needs an interactive coding agent connected to a local, OpenAI-compatible LLM endpoint (e.g., Ollama) to run its LLM skills.

Highlighted Details

• Generates multi-part technical tutorials from prompts.
• Features a local UI for learning with ToC, side-notes, and exercises.
• LLM skills enable Q&A, verification, and content extension.
• Tutorials document research sources, LLM model, and writing "voice."
• Supports custom LLM voices and includes built-in "plainspoken" and "companion" options.
• Offers opt-in tutorial verification for code execution correctness.
• Provides client-side search and filtering for tutorial libraries.

Maintenance & Community

Described as an experiment, primarily tested by the author on macOS with Claude Code. User feedback on other agents/OS via GitHub issues is encouraged. Specific community links or contributor details are absent.

Licensing & Compatibility

The repository's license is not stated, making commercial use or closed-source integration compatibility undetermined.

Limitations & Caveats

Generated tutorials may contain LLM hallucinations and lack human polish. Homebrew install is macOS-only. Full compatibility across all agents/OS is not guaranteed due to focused testing on the author's macOS setup. The absence of a clear license poses adoption risks.