Dugan Labs

Request pipeline middleware — security headers, CORS, rate limiting, CSRF, and adapters

Queue-driven local-inference primitive — drives a cloudflared-tunneled vLLM tower (or any OpenAI-compatible endpoint) with an SQLite-backed job queue, iteration loop runtime, fallback to Workers AI, and reusable job templates.

Eval harness measuring whether models actually produce correct BaseNative

Query builder with SQLite, PostgreSQL, and Cloudflare D1 adapters

Signal-based form state with dirty/touched tracking and schema validation adapters

ICU message formatting, locale detection, and @t template directive