Bridging the gap between fast-evolving code libraries and AI models

I’ve been working on an idea to make coding LLMs much smarter about new or less-known libraries. Right now, when developers use fresh packages or new versions, most AI coding tools fail multiple times because they don’t have context about those APIs—they end up searching the web and hallucinating.

My project aims to fix that by indexing packages and version, and exposing them through a unified API that LLMs (or AI agents) can query directly.

The system would understand which version the user’s environment uses, explain why a symbol is missing or deprecated, and suggest alternatives or migration paths.

Do you think this is a valuable direction for improving AI-assisted coding tools?