DocMeld CLI
Python CLI and library that turn PDF, Word and PowerPoint into structured elements, JSONL pages and skills with no OCR or models.
- 101 ★ on the repository
- Last commit
Extended access
Why this level
- The gold stage and knowledge generation send document text to the outside DeepSeek service.
- Processing reads and writes files in working folders.
Install
- Works in
In your terminal, with SkillFoxx CLI
npx skillfoxx add cli/docmeldDetects the agents on your machine, shows the access level and pins the version.
Other ways to install
Access and checks
How we reviewExtended accessThe tool runs code or makes network calls and only reads external data.
- The gold stage and knowledge generation send document text to the outside DeepSeek service.
- Processing reads and writes files in working folders.
This is third-party code. Review the repository files before installing.
What it does
DocMeld runs a document through three stages. Bronze turns the file into structured JSON with elements: titles, tables, images, formulas, speaker notes. Silver assembles self-contained JSONL pages with title hierarchy and Markdown tables, ready for search and RAG. Gold enriches pages with metadata through DeepSeek. The `process`, `bronze`, `silver` and `gold` commands are idempotent: a rerun skips files already processed. Since version 0.3.1 the `categorize`, `prd`, `workflow` and `skills` commands build topic clusters, a requirements document, a step-by-step workflow and Claude Code skill files from the text of a book or paper.
Who it is for. Developers of RAG and agent systems who need lightweight document parsing without heavy models and with a predictable element schema.
Good fit when
- A folder of PDFs and slide decks must become page-by-page JSONL for indexing.
- A set of Claude Code skills is wanted from a book.
- A description and keywords for each page are added in a separate pass.
- Papers in a folder must be sorted by topic with a `categories.json` file.
Not a fit when
- A scan recognizer is needed; DocMeld reads the text layer and has no OCR.
- Documents cannot go to an outside service; the gold stage and knowledge generation call DeepSeek.
- An MCP server is required; the repository has only a CLI and a library.
Example request
Process every PDF and PPTX in the reports folder with `docmeld process`, show the page and table counts, then generate skills from the book guide.pdf.Limitations
The bronze and silver stages work offline with no keys, while gold and the `categorize`, `prd`, `workflow` and `skills` commands need `DEEPSEEK_API_KEY` in `.env.local` and send page text to DeepSeek. The `.doc` and `.ppt` formats are read through LibreOffice, while `.docx` and advanced PDF parsing need extra packages (`docmeld[docling]`, `docmeld[pptx]`). Python 3.9 or newer is required. The last commit dates from September 2026 and the PyPI version is 0.3.1. The MIT license is stated in the README.
How to disable. Run `pip uninstall docmeld`. Processing results sit in folders beside the documents and are deleted by hand.
FAQ
How do I install DocMeld?
With `pip install docmeld`. Extras install like this: `pip install docmeld[docling]` for DOCX and advanced PDF, `docmeld[pptx]` for PowerPoint, `docmeld[office]` for everything.
Are models needed for parsing?
No. Bronze and silver work with no OCR, VLM or multimodal models, offline and without keys. AI is needed only for the gold stage and knowledge generation.
What is the gold stage?
An optional pass with retries on failure: an error on one page does not stop the others.
How do I get skills from a book?
With `docmeld skills book.pdf` or the `parser.process_skills()` method. `DEEPSEEK_API_KEY` is required and the result is Claude Code skill files.
Can another model be used?
Yes, knowledge generation works through the `LLMProvider` protocol, so the backend is replaceable. PDF parsers also switch with the `--backend` flag.
Discussion
No questions or reviews yet.
Related
CLI toolsget-md turns HTML pages, PDF, DOCX and Markdown into clean Markdown for language models, from the command line or code.
minirag-mcp searches your documents with hybrid vector and BM25 search and runs on your computer without sending files out
A CLI engine that compiles documents into a local structured wiki: a transparent alternative to plain RAG, reachable by an agent over MCP
A CLI and MCP server that extracts structured knowledge from documents: graphs, hypergraphs, and fact search with sources
Connects a Zotero library to Claude and other agents: search papers, annotate, cite, and search by meaning
An MCP server gives agents search and answers over local documents using hybrid retrieval, reranking and section links.
Comparison with similar tools
Cards that share the main search query, and how they differ.
| Tool | Official | Authentication | Repository stars | Main difference |
|---|---|---|---|---|
| DocMeld (this card) | no | 101 | ||
| Docling | no | 70,497 | a CLI that brings a mixed archive to a common format; DocMeld produces page-level JSONL with title hierarchy and generates skills | |
| MarkItDown | yes | not required | 195,183 | a Microsoft utility and MCP converting files to Markdown; DocMeld builds a three-stage pipeline and an element schema |
README in short
The project builds a bronze, silver, gold pipeline instead of plain format conversion: each stage runs separately and a schema of 14 element types is validated by Pydantic. The README has a comparison table against MinerU, Docling, Marker and MarkItDown.