Run

Execution model

  • Content-addressed outputs. Each run writes its result to .exuvia/artifacts/<node>/<sha>.parquet (or .json), keyed by the node's parameter set (see "Run parameters and replays"). Paths are stored relative to the project, so a project directory can be moved or restored elsewhere with its outputs intact.
  • Outputs in object storage (optional). Set EXUVIA_ARTIFACT_STORE to an S3 URL and the bucket becomes the durable copy of every output (AWS S3, Cloudflare R2, MinIO, anything S3-compatible). Each output is uploaded when its run finishes; the local .exuvia/artifacts/ folder becomes a cache that a fresh process refills on first read, and hub exports leave uploaded outputs out. Outputs recorded before the store was set (or whose upload failed) are uploaded in the background when the project loads, and stay in exports until they are. Deleting a project deletes its objects.
  • Freshness is derived, not flagged. A node is fresh when its code hash matches its last successful run and each upstream's current artifact is the one that run consumed. Otherwise it is stale, and the reason says why.
  • Run = target + stale ancestors, in dependency order, in parallel where the graph allows (EXUVIA_MAX_PARALLEL, default 4).
  • Subprocess isolation. Every node runs in its own Python process: crash isolation, clean per-run logs streamed live to the UI, killable timeouts (EXUVIA_RUN_TIMEOUT, default 1800 s).
  • Per-node schedules + cascade. A scheduled node is forced to re-run on its cron. If it produces different output, downstream nodes with cascade=True refresh automatically; identical output stops the cascade. Schedules can be paused from the UI without editing code.
  • Run history lives in a database: status, trigger (manual / schedule / cascade / agent / cli), timings, logs, schema. Schedule state (last fire per node) is persisted too, so restarts neither drop nor double-fire schedules. By default that is .exuvia/exuvia.db (SQLite) inside the project; see "Records: SQLite, Postgres, or your own adapter".
  • Every version of the code is kept. Each time a node file, a lib/ helper, a view file, exuvia.json or AGENTS.md changes, by any route (UI, API, agent, an editor, a template), the new text is saved as a revision with who saved it (agent, api or the caller's X-Exuvia-Origin, file, template, import, restore, load). A deletion is a revision too. Revisions are keyed by the same hash each run records, so every output can show the exact code that produced it, and any version can be restored (a restore is itself a revision). The agent has list_revisions, read_revision, restore_node and restore_view_file.