Server Execution Log
The execution log is used to view the run log of the remote workflow. It is located in the "execution log" workspace of the project details, and only shows the running of the released version or latest, and does not mix the draft records into the normal list.
Log Table
The log table reads run summaries by page and provides compact filtering:
- Search by execution ID, workflow name, session ID, target, or resolved target.
- Filter by All, In progress, Pending, Success, Failure, Cancel, or Timeout.
- Run source filtering by Desktop, External API, Embed, Webhook, session, or API Key.
- Displays the current page range and the total number filtered by the server.
- Using previous/next page paging, the next page is not appended to the current list.
The table displays the execution ID, status, target version, Workflow entry title/ID, creation time, and elapsed time. The status will not only be expressed by color, but also display the Chinese status badge. The Run ID is omitted when there is insufficient space, but the full title is left. The entry title comes from the snapshot at the time of run creation and is not renamed with subsequent drafts or versions; "Default/Legacy Entry" is displayed when the entry field is missing for the old record ".
When initiating a remote run using the API key generated by the Desktop login,Server will mark the run as external.source = "desktop". GET /api/workflows/{workflowName}/runs?source=desktop only returns such Desktop runs;CLI or plain manual API keys are not marked as Desktop sources.
row-level operations
Each log provides common operations:
| Operation | Description |
|---|---|
| Perform Playback | Read-only Run workspace playback is turned on; when Embed is enabled, the corresponding session deep chain is entered. |
| Detailed Log | Open the center pop-up box to view the stdout, stderr, runner diagnostics, and report.json. |
| Stop running | Show the stop entry for run on running or waiting and cancel the run after confirmation. |
During reading, stopping, and switching, the button is disabled and loading; when it fails, the error is displayed in the current page feedback area.
Playback
Perform playback to view the user-visible output of workflow:
- Normal workflow displays the exact resolved target, entry title/ID, parameter summary, status, and Output.
- conversation workflow summarizes transcript by the same
conversation_id. - Business timings, attachment entries, and output items are displayed in the same way as Desktop/Embed.
- The streaming body and tool results keep the Unicode characters intact and do not display replacement characters due to data segmentation.
- The streaming output of the public Embed only automatically follows when the view stays at the very bottom; the user actively scrolls up and then maintains the current position, returns to the bottom and resumes.
- Large args, report s, and output items are cropped for expandable previews to avoid rendering too much content at once.
Playback is a read-only view and does not provide a rerun, upload, or new session entry.
Waiting for input recovery will continue to use the exact resolvedTarget + entrypointId in the record, and the commit interface cannot switch entries. If draft deletes the original entry during the waiting period, the return 409 is restored, and the current default entry will not be used instead. If the old record has no entry field, it will be compatible with the default entry of the original precise target.
Detailed Log
The detailed log box is used to troubleshoot a single run record, including:
- Execution portrait: status, target, elapsed time, creation time, source, conversation, API Key, and execution user.
- Log summary: whether stdout, stderr,
report.jsonexist. - Searchable logs: stdout, stderr, runner diagnostics, and
report.json. - Read-only JSON viewer: Supports search targeting, previous and next matches.
stdout, stderr, runner diagnostics, and report.json start expanded and can each be collapsed manually. Every region fills the dialog's available content width; long text and JSON use bounded, dedicated scroll regions, and collapsing report.json leaves no fixed-height blank space.
If you want to view only the user-visible output, use the execution replay; if you want to troubleshoot the runner, node report, or stderr, use the verbose log.
Running status
| Status | Meaning |
|---|---|
running | Currently executing. |
waiting | Wait for user input or queue scheduling. |
success | Completed successfully. |
failed | Execution failed. |
aborted | Canceled. |
timed_out | Timeout or old run placeholder has expired. |
External Run API will map the result of success and errCode === 0 to succeeded; if workflow returns a non-zero errCode, it will be treated as a failure even if the process exits normally.
Mobile end and dark mode
The mobile table retains the readable state and can reach the operation button. The playback drawer occupies the available width, and the detailed log pop-up box uses the safe margin and the content area to scroll. In dark mode, status badges, filter controls, tables, drawers, and pop-up boxes remain recognizable.