ragmir-build
Installation
SKILL.md
Build with Ragmir
Ragmir is the retrieval layer: it parses, indexes, and searches documents on the machine that runs it and returns passages with exact citations (lines, pages, slides, cells, or sections). The application owns the model, the prompt, the answer, authentication, and where passages go. Ragmir calls no model and opens no port.
Choose the integration
| Stack | Use | Install |
|---|---|---|
| Node.js 22.12+ service or script | createRagmirClient() |
npm install @jcode.labs/ragmir |
| LangChain.js, LangGraph.js | RagmirRetriever, createRagmirTools from @jcode.labs/ragmir/langchain |
npm install @jcode.labs/ragmir @langchain/core |
| Python script, notebook, or service | Ragmir (sync, async twins), AsyncRagmir |
pip install ragmir |
| LangChain, LangGraph (Python) | RagmirRetriever, RagmirToolkit from langchain_ragmir |
pip install langchain-ragmir |
| LlamaIndex | RagmirRetriever from llama_index.retrievers.ragmir |
pip install llama-index-retrievers-ragmir |
| Any language with an MCP client | rgr serve-mcp over stdio |
the npm package or the Docker image |
| A host without Node.js | ghcr.io/jcode-works/ragmir, :full for semantic search and OCR |
Docker |