Execution detail

The deep-dive view for one execution. Reached via /logs/<id>, or by clicking a row in history or the log explorer.

Page structure

┌─────────────────────────────────────────────────────────────────┐
│ Header: Execution ID · Workflow name · Open in editor · Replay  │
├─────────────────────────────────────────────────────────────────┤
│ Tabs:  Overview  │  Output  │  Raw                              │
├─────────────────────────────────────────────────────────────────┤
│                                                                 │
│  Overview tab:                                                  │
│  ┌────────────┐ ┌────────────┐ ┌────────────┐                  │
│  │ State card │ │ Metrics    │ │ Cost card  │                  │
│  └────────────┘ └────────────┘ └────────────┘                  │
│                                                                 │
│  ┌─────────────────────────────────────────────────────────┐    │
│  │   Workflow canvas snapshot — coloured by node status    │    │
│  └─────────────────────────────────────────────────────────┘    │
│                                                                 │
│  Timeline table — one row per node                              │
│                                                                 │
└─────────────────────────────────────────────────────────────────┘

Header

• Execution ID with a Copy button.
• Workflow name with a link back to the editor (loads this run as a snapshot, read-only).
• Status badge (success / failed / …).
• Replay, Cancel (when applicable), Open in editor.

Overview tab

Workflow state card

• Execution ID
• Started at / ended at / duration
• Trigger type and input preview
• Workflow version that ran
• Worker / region that processed it (for distributed deployments)

Metrics card

• Total duration
• Node count (planned vs executed vs failed)
• Bytes processed
• API calls (count of HTTP nodes that fired)

Cost card

For runs that incurred cost (mostly AI):

• Total tokens (input + output)
• Total cost in USD
• Per-provider breakdown

AI metrics card (if applicable)

• Tokens by AI node
• Cost by AI node
• Cached calls
• Average response time

Canvas snapshot

A read-only render of the workflow with nodes coloured by their per-execution status:

Click a node in the snapshot to scroll the timeline below to its row.

Timeline table

One row per node that ran (or was skipped). Columns:

Expand any row to see the full JSON of input / output / error. Right-click on a field → Copy as expression.

Output tab

The final output of the workflow (whatever the Terminal node produced, or the last node's output if no Terminal). Two views: Pretty (formatted JSON) and Raw (the bytes).

For binary outputs (PDFs, images), shown as a download link.

Raw tab

Everything in one big JSON dump: state card data, every node's input / output / error, timings, metadata. Useful for sharing with support, "here's the full record of what happened".

Actions

• Replay, re-run the workflow with this same input. New execution ID; result lands in history.
• Cancel, for runs still in progress. Marks cancelled after the current node finishes.
• Open in editor, opens the workflow at the version that ran, with this run's data loaded into each node's inspector. Read-only. Helpful for tracing.
• Resolve approval, for runs in waitingForInput, jumps to the approval queue.
• Share, copy the URL. Anyone with workspace access can open it.

Tips & gotchas

• Open in editor loads a snapshot, not the live workflow. Changes you make in this mode aren't saved. To edit the live workflow, exit back to the regular editor URL.
• The canvas snapshot reflects the workflow version that ran. If you've since edited the workflow, the canvas in execution detail may not match what's in the editor right now.
• Large inputs / outputs are truncated in the row preview. Expand to see the full JSON; for very large payloads use the Raw tab.
• Cost numbers are estimates. They're computed at run time from the provider's per-token pricing table; if the provider changes pricing, historical estimates aren't retroactively updated.

Related

• Execution lifecycle
• Debugging failures
• Execution history
• Log explorer

Found something out of date? This page lives in the Flero docs content set.