Wires docToMarkdown/markdownToDoc into the editor clipboard: copying
selected content puts Markdown on text/plain alongside the browser's
own HTML (so pasting into a plain-text destination yields Markdown),
and pasting plain text that looks like a Markdown document converts it
to rich nodes; content with real HTML on the clipboard is left to
ProseMirror's normal HTML-based paste, and the heuristic requires two
or more distinct Markdown-shaped lines (or a fenced code block) so
ordinary prose is never mangled.
Both directions need the parsed/selected doc re-hydrated against
whichever schema instance is on the other side of the boundary: the
canonical editorSchema (packages/shared) for markdownToDoc's output
before inserting it into the live view, and the live view's schema
wrapped back into editorSchema before handing a slice to docToMarkdown
— they're structurally identical but not the same object, and
ProseMirror's content checks are identity-based.
Adds GET /pages/:id/export/markdown (downloads <slug>.md), serving the
already-derived page_content_cache.markdown (#23) rather than
re-decoding the Yjs state. "Copy as Markdown" and "Download as
Markdown" actions in the page header both read from that same
endpoint, so they always agree with each other and with the last saved
state.
Closes#30
