Documentation
Execution controls: retries, timeouts and error handling
Updated September 17, 2026
Real workflows call models and APIs that fail. Novalink gives every node its own failure policy and every run a budget, so one slow call cannot stall or overspend a whole workflow.
Per node
- Timeout: 120 seconds by default.
- Attempts: 1 by default, meaning no retry. Retries wait with backoff, starting at 2 seconds.
- On error: what happens after the last attempt fails, described below.
Error modes
stop(default): the run fails.continue: the node fires no ports, so everything downstream of it is skipped while the rest of the run carries on.error_port: the node fires itserrorport with the message, error type and attempt count, so you can route failures to a fallback branch.
Per run
max_parallel: independent branches run concurrently, 4 at a time by default.run_timeout_seconds: 900 by default.max_total_tokens: a cost guard across every agent in the run, 200,000 by default.
Cancelling a run
While a run executes, the Run button in the toolbar becomes Cancel. Every event up to that point is kept.
Keep reading
- Building workflows on the canvasHow Novalink workflows are structured: triggers, nodes, ports, data edges and tool edges, branching and merging, and how the executor decides what runs.
- Designing a workflow engine you can watch runHow Novalink's executor works: one rule for active and dead edges, run events streamed over SSE, and graph snapshots that make run replay trustworthy.
- Merge node: join branchesThe Merge node joins workflow branches back together once every incoming edge settles, passing on the first value, a list, or an object keyed by source.
Build it on the canvas
Create a free account and run your first workflow in the browser, or run Novalink on your own machine.