@kuralle-agents/rag-loaders

Document loaders for Kuralle RAG pipelines: PDF, URL, CSV, and Markdown.

Install

npm install @kuralle-agents/rag-loaders @kuralle-agents/rag

Peers: @kuralle-agents/rag. Optional per-loader: pdf-parse (PDF), cheerio (URL), papaparse (CSV).

What it does

Four DocumentLoader implementations that return Document[] for indexing into any Kuralle vector store or RAG pipeline. An opt-in registry dispatches by file extension.

Key exports:

• PdfLoader — load a local PDF file or in-memory buffer.
• UrlLoader — fetch and parse an HTTPS page.
• CsvLoader — load a CSV file as structured documents.
• MarkdownLoader — load a Markdown file, splitting by headings.
• registerLoader / loadForPath — opt-in registry: dispatch by extension.

Usage

import { PdfLoader, MarkdownLoader } from '@kuralle-agents/rag-loaders';

const pdfDocs  = await new PdfLoader({ filePath: './policy.pdf' }).load();
const mdDocs   = await new MarkdownLoader({ filePath: './guide.md' }).load();

Loader registry

Register only what you import; unused loaders stay tree-shakeable.

import { registerLoader, loadForPath, PdfLoader, MarkdownLoader } from '@kuralle-agents/rag-loaders';

registerLoader('pdf', (path) => new PdfLoader({ filePath: path }));
registerLoader('md',  (path) => new MarkdownLoader({ filePath: path }));

const docs = await loadForPath('./whitepaper.pdf');

Size limits

| Loader | Practical limit | |--------|----------------| | PdfLoader | ≤ 25 MB — pdf-parse loads the full document into memory. | | UrlLoader | ≤ 5 MB body (configurable via maxBytes). | | CsvLoader | ≤ 100 MB — full Document[] materialized in memory. | | MarkdownLoader | ≤ 50 MB — full-document load. |

All loaders return Promise<Document[]>; streaming is not supported.

Related

• @kuralle-agents/rag — chunkers, retrievers, vector stores.
• @kuralle-agents/tools — createVectorRetrievalTool.