How wrtnlabs/autoview works
Project type and purpose
• This is a PNPM-powered monorepo hosting an open-source developer tool plus companion libraries and website.
• Core purpose: “Autoview” automatically generates React viewer components from TypeScript or OpenAPI schemas using an LLM + typia validation + a code‐generation pipeline.
File structure
• Top-level “deploy”, “build”, CI workflows (.github), and a “packages” folder.
• “packages” contains:
• agent: orchestrates schema ingestion, LLM calls, and codegen
• compiler & compiler-bundle/browser-test: internal code-generation engine, bundlers, and test harnesses
• interface: TypeScript interfaces & metadata services for schema descriptions
• ui: runtime React/Tailwind/shadcn/ui library plus a component renderer
• “research” and “test” folders contain data‐driven experiments and automated test suites.
• “website” is a Next.js app for docs and demos.
Key insights from README
• LLM Function Calling + typia schema analysis
• Generates TypeScript/TSX code for React viewer components
• Depends on OpenAI (or other LLM vendors) for code composition
• Output is consumed by @autoview/ui’s renderComponent