Documentation
The docs are Markdown in docs/, built with Zensical
and published at https://neuroseq.github.io/. Navigation is in zensical.toml.
Layout
| Folder | Content |
|---|---|
docs/getting_started/ |
Installation, quickstart, GUI overview, chat assistant |
docs/networks/ |
Simulators, external backends, hardware substrates |
docs/input_design/ |
Static and dynamic input design, export |
docs/readout/ |
Experiments, imported imaging and patch-clamp data |
docs/data_analysis/ |
Analyses, plugins, simulation vs. experiment |
docs/developer/ |
This guide |
docs/images/ |
Images |
A new page must be added to the nav in zensical.toml.
Docs feed the chat assistant
The wheel bundles docs/ as neuroseq/docs ([tool.hatch.build.targets.wheel.force-include]
in pyproject.toml). neuroseq.llm.system_prompt concatenates every .md file
into the assistant's system prompt, from the installed package or, in a checkout,
from the repository's docs/. Keep pages short and accurate: every word costs
prompt tokens and stale text produces wrong answers.
Build and preview
uv run zensical serve # live preview at the URL shown
uv run zensical build --clean --strict # what CI runs; output in site/
With Conda, install the docs group (python -m pip install --group docs) and
drop uv run. prek run -a formats Markdown with prettier.
Deployment
.github/workflows/docs.yml runs the strict build on pull requests and pushes
to main, and can be started manually. Pushes to main publish site/ to
neuroseq/neuroseq.github.io,
so the site lives at neuroseq.github.io rather than neuroseq.github.io/neuroseq.
With a custom domain, that extra repository and its deploy key would not be needed.