Hey Product Hunt! 👋 I built askable-ui because every AI copilot I tried had the same problem: it couldn't see what the user was looking at. The user had to describe it. That's friction, and it's lossy. The idea is stupid simple: annotate UI elements with `data-askable`, and the library tracks focus + serializes it into a prompt-ready string. Pass that string to any LLM — OpenAI, Anthropic, Vercel AI SDK, CopilotKit, LangChain. That's it. Your assistant now knows the user is focused on Q3 revenue. What's inside: ✅ Zero runtime dependencies ✅ React, Vue 3, Svelte, React Native, Vanilla JS ✅ SSR-safe ✅ Viewport awareness (what's currently on screen) ✅ Conversation history helpers for multi-turn ✅ Redaction hooks for sensitive fields ✅ Live inspector overlay for dev Not replacing your LLM SDK — askable-ui is the context layer underneath it. Think of it as the "eyes" layer. MIT licensed, open source, and I'd love feedback — especially on the Vue and Svelte adapters. Drop questions in the comments!

askable

Two lines of code to give your LLM eyes.

Two lines of code to give your LLM eyes.