zenodo-publish

Markdown to Zenodo DOI in one command — with AI-powered multi-agent peer review.

python publish.py --file paper.md --title "My Paper" --description "Abstract here"
# → DOI: 10.5281/zenodo.XXXXXXX

What it does

1. Converts your Markdown file to a styled PDF
2. Creates a deposit on Zenodo via the REST API
3. Uploads the PDF
4. Sets metadata (title, abstract, keywords, license, related works)
5. Publishes and returns the DOI

Your paper is immediately citable, indexed by DataCite, and (if linked) auto-harvested by ORCID.

Setup

git clone https://github.com/Life-Atlas/zenodo-publisher.git
cd zenodo-publisher
pip install -r requirements.txt
cp .env.example .env
# Edit .env with your Zenodo token

Get your token at zenodo.org/account/settings/applications. Scopes needed: deposit:write + deposit:actions.

Usage

New publication

python publish.py \
  --file paper.md \
  --title "The Reality Construct" \
  --description "Position paper on digital twins as boundary-spanning objects." \
  --keywords "digital twins,boundary objects,actor-network theory"

Dry run (creates draft, doesn't publish)

python publish.py --file paper.md --title "Test" --dry-run

Update an existing deposit (new version)

python publish.py \
  --file paper-v2.md \
  --title "The Reality Construct (v2)" \
  --update 19586835

The original DOI continues to resolve. The new version gets its own DOI. Both are linked.

Link related works

python publish.py \
  --file paper.md \
  --title "Domain Paper" \
  --related "10.5281/zenodo.19587944:continues,10.5281/zenodo.19586851:references"

Relations: continues, isPartOf, references, isSupplementTo, isCitedBy, isNewVersionOf, etc. (full list).

Custom creator

python publish.py \
  --file paper.md \
  --title "Paper" \
  --creator "Doe, Jane:MIT:0000-0000-0000-0001"

Format: Last, First:Affiliation:ORCID (affiliation and ORCID optional).

Sandbox (testing)

python publish.py --file paper.md --title "Test" --sandbox

Uses sandbox.zenodo.org — separate token needed from sandbox.zenodo.org/account/settings/applications.

All options

Flag	Description	Default
--file	Path to Markdown file	required
--title	Publication title	required
--description	Abstract (HTML supported)	title
--keywords	Comma-separated keywords	none
--type	workingpaper, article, preprint, report	workingpaper
--date	Publication date (YYYY-MM-DD)	today
--creator	Last, First:Affiliation:ORCID	see .env
--related	DOI:relation pairs, comma-separated	none
--update	Deposit ID to create new version of	none (new)
--funding	Zenodo grant IDs, comma-separated	none
--communities	Zenodo community IDs, comma-separated	none
--dry-run	Create draft but don't publish	false
--sandbox	Use sandbox.zenodo.org	false

PDF quality

The tool tries two PDF backends:

1. WeasyPrint (recommended) — styled HTML-to-PDF with proper typography. Needs system libraries (libpango, libcairo). Native on Linux/Mac, requires GTK for Windows.

2. ReportLab (fallback) — pure Python, works everywhere, basic formatting. Install with pip install reportlab.

AI Peer Review

Before publishing, run your article through 5 parallel AI reviewers:

# Standalone (requires ANTHROPIC_API_KEY)
python review.py --file paper.md --threshold 9.0

# Output:
#   STS / Theoretical Rigour: 8.5/10
#   Engineering / Technical:  8.0/10
#   IS / Management:          9.0/10
#   Domain Expert:            8.5/10
#   Quality / APA:            9.0/10
#   
#   COMPOSITE: 8.6
#   RATING: MINOR REVISION
#   ✗ BELOW threshold (9.0). Revision needed. Gap: 0.4 points

Five reviewer personas score across weighted dimensions:

Dimension	Weight	What It Measures
STS / Theoretical Rigour	25%	Frameworks applied, not just cited
Engineering / Technical	20%	Claims supported, architecture coherent
IS / Management	15%	Benefits-driven, falsifiable methodology
Domain Relevance	20%	Practitioners would use this
Quality / APA	20%	Formatting, references, tone

The review → fix → re-review loop typically improves articles from 5-6 (first draft) to 8.5+ within 3-4 rounds.

See SCORING.md for the full rubric and CLAUDE_CODE_USAGE.md for Claude Code integration.

# Save full review as JSON
python review.py --file paper.md --output review.json

# Use a different model
python review.py --file paper.md --model claude-opus-4-6

GitHub Actions

To auto-publish on release:

# .github/workflows/publish.yml
name: Publish to Zenodo
on:
  release:
    types: [published]
jobs:
  publish:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
      - uses: actions/setup-python@v5
        with:
          python-version: '3.12'
      - run: pip install -r requirements.txt
      - run: |
          python publish.py \
            --file ${{ github.event.release.body }} \
            --title "${{ github.event.release.name }}"
        env:
          ZENODO_TOKEN: ${{ secrets.ZENODO_TOKEN }}

License

MIT

Author

Nicolas Waern — WINNIIO AB — ORCID

Built as part of the Life Atlas research programme.