Workflows for Coda, deposited by the people who use it.
Open Coda and pick Examples ▸ Browse Workflows… to search this repository from inside the app; clicking an entry loads it onto the canvas. Nothing here needs to be downloaded by hand.
A workflow worth depositing is one that answers a question somebody else also has, and that carries enough explanation to be adapted rather than only re-run. A three-node graph with a good README beats a twenty-node one without.
Most entries need a token for the connectome they query — Coda keeps credentials in Connections and none of them ever travel inside a graph. Each entry declares what it needs, and the browser shows that on the card before you open it.
See CONTRIBUTING.md. The short version:
workflows/
your-workflow-slug/
graph.coda.json what Coda's Save button writes
meta.json name, summary, tags, authors, requires
README.md what it does, why the steps are in that order
then open a pull request. index.json is generated — CI regenerates and commits it.
index.json is the only file Coda downloads to build the list — one request to
raw.githubusercontent.com, which is CORS-open and caches for five minutes. The graph itself is
fetched when you open an entry.
It is generated, not hand-edited, and CI checks that it matches workflows/. The generator
lives in Coda (scripts/zoo-index.ts) rather than here because validating a deposited graph
means loading Coda's node registry: whether a node type still exists, whether a param survived a
rename, whether the declared requirements match the dataset nodes in the graph. A weekly
scheduled run re-validates everything against current Coda, so a workflow that Coda has drifted
away from shows up as a failed build here rather than as a broken card in somebody's browser.