A DeepSeek Harness Cordis plugin that exposes the CazzPatent Python toolchain as five deterministic, schema-validated tools. Pair it with the cazz-patent skill (the instruction brain) for the full experience, or use the tools on their own.
| Tool | Wraps | Purpose |
|---|---|---|
patent_md_to_docx |
md_to_docx.py |
Markdown → Word (.docx), LaTeX→OMML + CJK + tables |
patent_docx_to_pdf |
docx_to_pdf.py |
Word → PDF (Word COM / LibreOffice / weasyprint) |
patent_render_diagram |
html_to_png.py |
HTML/SVG → high-res PNG (3× DPI) |
patent_batch_diagrams |
batch_diagrams.py |
Batch Mermaid → HTML+SVG+PNG |
patent_html_to_pdf |
html_to_pdf.py |
Markdown/HTML → PDF (weasyprint / Playwright) |
Every tool returns a uniform canonical value: { exitCode, timedOut, aborted, stdout, stderr }.
| Field | Default | Meaning |
|---|---|---|
pythonPath |
python |
Python interpreter used to invoke the scripts. |
scriptsDir |
bundled scripts/ |
Directory containing the Python scripts. |
defaultTimeoutMs |
300000 |
Foreground timeout for every run. |
Override in a profile's cordis.patch.yml (or edit this bundle's cordis.patch.yml):
- insert:
- id: cazz-patent-tools
name: dsh-cazz-patent
config:
pythonPath: D:/miniforge3/envs/patent/python.exe
scriptsDir: /absolute/path/to/CazzPatent/cazz-patent/scriptsThe plugin ships as an npm bundle (its package.json declares dsh.bundle). Install from a git host:
dsh plugin --profile demo add github:you/CazzPatent#<sha>
# pnpm will ask you to allow the build (the prepare script compiles src/ -> lib/):
# copy the printed package key into the profile's pnpm-workspace.yaml allowBuilds, then re-run.
# Then boot:
dsh --profile demoOr from a local checkout during development:
dsh plugin --profile demo add ./pluginThe wrapped scripts need Python 3.10+ and the CazzPatent dependencies. Install them into the interpreter referenced by pythonPath:
pip install -r ../requirements.txt
playwright install chromium
# Windows DOCX → PDF (highest fidelity)
pip install pywin32pnpm install
pnpm run build # esbuild bundles src/ -> lib/The prepare script runs the same build automatically after a git install. esbuild is the only build dependency (no DSH-internal packages), so the build is fully self-contained; @deepseek-ai/* imports are marked external and resolve from the dsh installation at runtime.