Documentation search for a small language is only half useful — the examples run out. Here the build and run tools close the loop: code is written, compiled, and the compiler's errors come straight back, so what you are handed has been through the compiler rather than around it. Crawling is not exposed as a tool, so nothing in a conversation can trigger a re-index.
A server for the Onyx programming language. It searches the official documentation and indexed GitHub repositories for examples, function definitions and struct definitions, and it compiles and runs Onyx code so compiler errors come back as output rather than guesses.
- Official documentation searched
- GitHub code examples searched by topic, plus function and struct definitions pulled from those repositories
- The list of repositories that have been indexed
- One search across every source at once
- Onyx code executed in a sandbox and its output or errors returned
- Code built in a project directory you name, and a package built the same way
- WebAssembly executed and its output returned
- A read-only interface — crawling is a command-line operation and is deliberately not exposed as a tool
The Onyx compiler installed and on the path for anything that builds or runs code, and Node for the server. Connect to the maintained hosted instance and there is nothing to install or index; run it locally and you populate the index yourself from the command line.