pyscimd 0.1.31

SciMD — Scientific Markdown for the AI Era

SciMD — Scientific Markdown for the AI Era

An open document format designed for humans to write, machines to understand, and science to advance.

---

The Problem

Modern scientific and technical documents are trapped in formats designed for printing, not understanding:

• PDFs require OCR to extract text, losing structure and introducing errors
• Charts as images are opaque to AI — a bar chart becomes meaningless pixels
• Figures without context force models to hallucinate interpretations
• Unstructured text makes RAG retrieval imprecise and chunk boundaries arbitrary
• Formulas as images cannot be parsed, searched, or validated

The result: AI models hallucinate, RAG systems retrieve noise, and training pipelines waste compute on lossy format conversions.

The Solution

SciMD (.smd) is a plain-text, human-readable document format that solves these problems by design:

Problem	SciMD Solution
OCR errors	Plain text with Markdown — no conversion needed
Opaque charts	Author provides tabular data + interpretation
Ambiguous figures	Mandatory author descriptions for every image
Poor RAG chunking	Semantic sections with unique IDs and metadata
Formula images	Native LaTeX inline ($...$) and block ($$...$$)
Diagram ambiguity	MermaidJS source code + author description
Missing context	Structured metadata at document and section level
Training noise	Sequential, predictable structure from authoring

Key Principles

1. Author-time structure — Structure is defined when writing, not reverse-engineered later
2. Data over pixels — Charts are data tables; diagrams are code; formulas are LaTeX
3. Interpretation is mandatory — Every visual element carries the author's explanation
4. Sequential readability — Documents flow logically for both humans and token streams
5. Plain text first — Every .smd file is valid UTF-8 text, editable in any text editor
6. Open by default — MIT licensed, community-driven, no vendor lock-in

Quick Example

---smd
title: "Effects of Temperature on Catalyst Performance"
authors:
  - name: "María García"
    orcid: "0000-0002-1234-5678"
    affiliation: "UNAM"
version: "0.1.0"
lang: "es"
keywords: ["catalysis", "temperature", "kinetics"]
---

::section{#intro}
::meta
type: introduction
summary: "Overview of temperature effects on heterogeneous catalysis"
::

# Introduction

The relationship between temperature and catalytic activity follows
the Arrhenius equation $k = A e^{-E_a / RT}$, where $E_a$ is the
activation energy and $R$ is the gas constant.

::endsection

::section{#results}
::meta
type: results
summary: "Experimental measurements of conversion rates at 5 temperatures"
depends_on: ["#methods"]
::

# Results

::chart{#fig-conversion}
::interpretation
Conversion rate increases linearly between 200–350°C, reaching a
plateau at 89% above 400°C. The inflection point at 350°C suggests
a change in the rate-limiting step.
::
| Temperature (°C) | Conversion (%) | Selectivity (%) |
|---|---|---|
| 200 | 23.1 | 95.2 |
| 250 | 41.7 | 93.8 |
| 300 | 62.4 | 91.1 |
| 350 | 78.9 | 87.3 |
| 400 | 89.2 | 82.6 |
::endchart

::endsection

Project Structure

scimd/
├── spec/
│   └── SPECIFICATION.md    # Full format specification (v0.1.0)
├── examples/
│   ├── basic.smd           # Simple example document
│   └── full-paper.smd      # Complete scientific paper
├── parser/
│   ├── scimd_parser.py     # Reference parser in Python
│   ├── scimd_validator.py  # Validation tool
│   └── requirements.txt    # Python dependencies
├── docs/
│   ├── AUTHORING_GUIDE.md  # How to write SciMD documents
│   └── RAG_GUIDE.md        # How to use SciMD for RAG pipelines
├── LICENSE
├── CONTRIBUTING.md
└── README.md

Getting Started

Writing a Document

Any text editor works. Save your file with the .smd extension and follow the Authoring Guide.

Validating a Document

pip install scimd
scimd validate my-paper.smd

Parsing for RAG

from scimd import SciMDParser

doc = SciMDParser.parse("my-paper.smd")

# Get semantically chunked sections
for section in doc.sections:
    print(section.id, section.summary)
    print(section.content)

# Extract all chart data as DataFrames
for chart in doc.charts:
    print(chart.interpretation)
    print(chart.dataframe)

For the Scientific Community

SciMD is built to serve researchers, not platforms:

• No proprietary tools required — write in VS Code, Vim, Notepad, anything
• Version control friendly — plain text diffs cleanly in Git
• Citation ready — structured metadata maps to BibTeX, CSL, and DOI
• Multilingual — UTF-8 native, lang metadata per document and section
• Accessible — mandatory descriptions make content accessible by design

Roadmap

• v0.1.0 — Core specification
• v0.2.0 — Reference parser + validator (Python)
• v0.3.0 — VS Code extension with live preview
• v0.4.0 — Pandoc filter for PDF/HTML/DOCX export
• v0.5.0 — LLM training pipeline toolkit
• v1.0.0 — Stable specification after community review

Contributing

We welcome contributions from researchers, developers, and anyone who cares about making knowledge more accessible. See CONTRIBUTING.md.

Author

SciMD was created by Juan Francisco Avilés Calderón.

License

MIT — Use it, fork it, improve it.

---

SciMD: Because science deserves better than screenshots of spreadsheets.