to-markdown

@cognipeer/to-markdown · Utilities

stable

A TypeScript-first utility that converts 20+ formats — PDF, DOCX, PPTX, XLSX/CSV, HTML, Jupyter notebooks, ZIP archives, RSS/Atom, images, and audio — into clean Markdown, ready for RAG pipelines and LLM context windows.

Quickstart GitHub Full docs

License · MITTypeScript · 100%Audience · Anyone building a…Fits · RAG ingestion

RAG ingestionDocument normalisationTypeScript-firstBrowse recipes

Production install

$npm install @cognipeer/to-markdown

What's in the box

Anyone building a RAG pipeline or document workflow that needs to normalise heterogeneous file formats into a single, LLM-friendly representation. Each capability is opt-in — use the parts that fit, leave the rest.

20+ formats

PDF, DOCX, PPTX, XLSX, CSV, HTML, IPYNB, ZIP, XML/RSS/Atom, images, and audio — all through one convert call.

Promise-based API

`convertToMarkdown(input)` takes a path, Buffer, or base64 string and returns a Markdown string with auto format detection.

TypeScript first

Written in TypeScript with full type definitions and zero-config import in modern toolchains.

AI-ready rich output

`convertToRichMarkdown` adds frontmatter, a section tree, and optional token-aware RAG chunks for ingestion.

Optional OCR

Opt-in image and scanned-PDF OCR via the tesseract.js peer dependency, with auto / always / never PDF modes.

Modular & fast

Per-format converters keep the bundle lean — only load the adapters you actually need.

How it runs

A small, modular pipeline: detect the source format, run the right adapter, normalise the structure, and emit clean Markdown — ready to chunk and embed.

Step 1

Detect

Format sniff

Step 2

Adapter

PDF · DOCX · HTML …

Step 3

Normalise

Headings · tables

Step 4

Emit

Markdown string

Inputs

PDF · DOCX
HTML · plain text
Excel · CSV
Bring-your-own buffer

Options

imageMode
tableStrategy
headingDepth
chunkHint

Output

Clean Markdown
Stable headings
Per-doc metadata
Streaming-friendly

Quickstart

Install, configure, run. The example below is the smallest piece of code that does something useful in production.

import  { convertToMarkdown, convertToRichMarkdown } from "@cognipeer/to-markdown";
 
// Path, Buffer, or base64 — format auto-detected
const md = await convertToMarkdown("./policies/handbook.pdf");
console.log(md);  // clean Markdown string
 
// AI-ready: frontmatter, sections, and RAG chunks
const doc = await convertToRichMarkdown("./report.docx", {
  includeChunks: true,
  chunkOptions: { maxTokens: 800 },
});
console.log(doc.chunks, doc.sections);

How it compares

Against the utilities options teams most often weigh — focused on operational concerns, not feature inventories.

Capability	to-markdown	pandoc	unstructured.io	turndown
PDF	● native	◐ partial	● native	○ missing
DOCX / PPTX	● native	● native	● native	○ missing
HTML	● native	● native	● native	● native
Excel / CSV	● native	◐ partial	● native	○ missing
TypeScript	● native	○ missing	○ missing	● native
Open source	● native	● native	◐ partial	● native