RAG and semantic search,
self-hosted in minutes
Add semantic search and RAG to your existing app in minutes. Self-hosted on your own infra — bring your own AI with Ollama or OpenAI.
$ npx ragpack init && npx ragpack start
✔ API on :9000 · Admin UI on :3000
# Drop into any existing app
import { RagPack } from "ragpack-js";
const client = new RagPack({
baseUrl: "http://localhost:9000",
});
const col = client.collection("my-docs");
await col.ingest({
uri: "https://your-docs-site.com/guide",
});
const { answer } = await col.rag({
query: "how does auth work?",
});Lightweight by design
The Go backend idles at ~20MB RAM. A single static binary with no runtime dependencies means you can run a full RAG pipeline on the smallest machine your cloud provider offers — not a dedicated $200/mo AI instance.
You choose the model
Ollama for fully local embeddings, OpenAI, or HuggingFace TEI. Switch providers by changing one env var — no code changes, no re-ingestion. Keep costs down with a local model or scale up when you need it.
Your data, your infra
Everything runs in Docker. Vectors and metadata live on your own disk. No usage-based billing, no rate limits imposed by a third party, no vendor to negotiate with as you scale.
What's included
REST API
Collections, ingest, query, and document management — all via a clean REST API.
JS / TS SDK
Typed client for Node and the browser. Drop RAG into your app in a few lines.
Admin UI
Built-in Next.js interface to manage collections, monitor jobs, and run queries.
Multiple embedding providers
Ollama, OpenAI, and HuggingFace TEI supported out of the box.
Supported formats
Markdown, HTML, PDF, and plain text — via URL, S3, or file upload.
Chunking strategies
Format-aware chunking out of the box. Configurable chunk size and overlap.
What can you build?
Docs chatbot
Let users ask questions about your documentation and get answers grounded in your content.
Knowledge base search
Make internal wikis, SOPs, and runbooks searchable with natural language — not just keywords.
AI customer support
Ground your support chatbot in your own product docs so it only answers from what you've written.
Code search
Ingest your codebase and let developers find relevant files, functions, or patterns semantically.