Skip to content

Quick start

This page assumes you have installed doctopus-ai.

Parse a document

doctopus parse path/to/document.pdf

Markdown goes to stdout and a short routing summary goes to stderr, so you can pipe the output:

doctopus parse paper.pdf > paper.md
doctopus parse paper.pdf --format json --output paper.json

See why an engine was chosen

plan runs inspection and routing without extracting anything:

doctopus plan paper.pdf
doctopus plan paper.pdf --mode accurate

The three processing modes are fast, balanced (default) and accurate. See Routing & fallback.

Pick an engine

doctopus backends list                       # what is installed and why others are not
doctopus parse report.docx --backend docling

From Python

import doctopus

doc = doctopus.parse("paper.pdf")

print(doc.markdown)
print(doc.tables)
print(doc.processing.backend)
print(doc.processing.decision.reasons)

chunks = doctopus.chunk("paper.pdf", max_chars=1500)

Build a RAG corpus

doctopus chunk ./corpus --output corpus.jsonl

Each line is one chunk with its heading path, pages and source. See Chunking.

Serve it

doctopus serve --port 8000

Open localhost:8000 for the drag-and-drop demo, or /docs for the interactive API documentation. See REST API & demo UI.

Next steps

  • Engines: what each backend is good at and what it costs.
  • MCP server: give an agent safe access to a document folder.
  • Output schema: every field in the Document.