DEV Community

Cover image for 116 Pre-Built Documentation Packages for Your AI Coding Assistant
Moshe Simantov
Moshe Simantov

Posted on • Originally published at neuledge.com

116 Pre-Built Documentation Packages for Your AI Coding Assistant

#documentation #productivity #showdev #tooling

Every time someone set up @neuledge/context for a new project, they'd do the same thing: clone the React docs repo, find the right directory, build a package. Then do it again for Next.js. And Tailwind. And Prisma.

I kept seeing the same repos show up in GitHub traffic. Hundreds of developers, all independently building identical documentation packages for the same popular libraries. That felt like a problem worth solving.

So we built a community registry — a shared collection of pre-built documentation packages that anyone can download instead of building from source.

The problem was simple repetition

The context add workflow works great. You point it at a repo, it finds the docs, builds a searchable SQLite database. Done.

context add --name react https://github.com/reactjs/react.dev /src/content/reference
Enter fullscreen mode Exit fullscreen mode

But for popular libraries, this is redundant work. You need to know the right repo URL, find the correct docs directory (which sometimes takes a few minutes of browsing), and wait for the build. Multiply that by every developer who uses React, and it's a lot of collective time spent producing the exact same .db file.

The registry just short-circuits that. Someone builds the package once, and everyone else downloads it.

What's actually in a registry package

Same thing you'd get building locally — a SQLite .db file with FTS5 full-text search, containing semantically chunked documentation. There's no difference between a package you build yourself and one from the registry. Same format, same search quality, same everything.

Right now the registry has packages for three ecosystems:

  • npm (109 packages): React, Next.js, Angular, Vue, Svelte, Astro, Tailwind CSS, Express, Fastify, NestJS, Prisma, Drizzle, and a lot more
  • pip (4 packages): Django, FastAPI, Flask, Pydantic
  • maven (3 packages): Spring Boot, JUnit, Micrometer

Packages get rebuilt daily through GitHub Actions. When Next.js ships a new version, the registry picks it up automatically. No one needs to do anything.

How to use it

The simplest way is context install:

context install npm/react
Enter fullscreen mode Exit fullscreen mode

That downloads the pre-built package and makes it available to your AI coding assistant immediately. If you want a specific version:

context browse npm/react        # see what's available
context install npm/next 15.0   # install a specific version
Enter fullscreen mode Exit fullscreen mode

Or with npx if you haven't installed @neuledge/context globally:

npx @neuledge/context install npm/react
Enter fullscreen mode Exit fullscreen mode

If you're running @neuledge/context as an MCP server, your AI agent can also find and install packages on its own. It has two tools for this — search_packages to find what's available, and download_package to install it. So if it encounters a library it doesn't have docs for, it can just go grab them from the registry without you doing anything.

How the registry works

The pipeline is pretty straightforward:

  1. Registry entries are YAML files that map a package name to a git repo and docs path. Each one says "for this library, clone this repo, look in this directory."

  2. A daily GitHub Actions workflow checks for new library versions. When it finds one, it clones the repo, builds the documentation package, and publishes it to the registry API.

  3. The API at api.context.neuledge.com serves search and download endpoints. Search to find packages, download to get the .db file.

  4. The packages themselves are the same SQLite databases @neuledge/context uses locally — meta table for metadata, chunks table for the documentation sections, and chunks_fts for full-text search.

If you want to add a library that's missing, you submit a YAML file to the GitHub repo with the package mapping. The build pipeline handles everything else from there.

What's covered

A quick overview of the categories:

Frontend: React, Next.js, Angular, Vue, Svelte, SvelteKit, Astro, Solid, Remix, Nuxt, Gatsby

CSS: Tailwind CSS, Sass, PostCSS, Styled Components, Emotion

Backend: Express, Fastify, NestJS, Hono, Django, FastAPI, Flask, Spring Boot

Database/ORM: Prisma, Drizzle, TypeORM, Mongoose, Sequelize, Knex

Testing: Jest, Vitest, Playwright, Cypress, Testing Library

AI SDKs: OpenAI SDK, Anthropic SDK, LangChain, Vercel AI SDK

Build tools: Vite, Webpack, esbuild, Turbo, Bun, Deno

Infra: Docker, Kubernetes, Terraform, AWS CDK

If you're working with a typical stack — say React, Next.js, Prisma, and Tailwind — that's four install commands and your AI assistant has accurate, version-specific docs for everything.

Try it out 

npx @neuledge/context install npm/react
Enter fullscreen mode Exit fullscreen mode

Or set up @neuledge/context as an MCP server and let your AI agent discover packages on its own. The getting started guide walks through the full setup.

The registry is open source and free. If your favorite library isn't there yet, adding it is one YAML file.

Top comments (0)