The source of https://rerun.io/docs and every code snippet it shows are plain files. Read them instead of guessing.
Where
- In a Rerun checkout:
docs/. - Inside the Rerun Viewer's agent panel: unpacked next to the skills, at
../../../docsrelative to this file. The session preamble names the absolute path.
Layout
| Path | What |
|---|---|
docs/content/**/*.md |
The website pages. docs/content/concepts/<name>.md maps to rerun.io/docs/concepts/<name>. |
docs/content/reference/types/ |
One page per archetype, component, view, and blueprint type. |
docs/content/reference/viewer/, reference/sdk/, reference/cli.md |
Viewer, SDK, and CLI reference, including the viewer-mcp server. |
docs/content/howto/, docs/content/getting-started/ |
Task-oriented guides and tutorials. |
docs/content/changelog/ |
Release notes, oldest to newest. |
docs/snippets/INDEX.md |
Table of every snippet: feature, name, one-line description, and which languages exist. Start here. |
docs/snippets/all/<category>/<name>.{py,rs,cpp} |
The snippets themselves. archetypes/, views/, howto/, concepts/, tutorials/, quick_start/. Same name in each language means the same example. |
Each markdown page starts with front matter (title, order); the body is the page.
Recording files (*.rrd) next to the snippets are not shipped with the viewer.
How
- Search first, then read:
rg -l "<term>" docs/content docs/snippets/all. - For "how do I log X": open
docs/snippets/all/archetypes/<x>.py(or.rs,.cpp) anddocs/content/reference/types/archetypes/<x>.md. - For "how does X work":
docs/content/concepts/. - When pointing the user at a page, give the website URL derived from the file path, not the local path.
- The Python API reference and
docs.rs/rerunare not on disk; useWebFetchfor those.