Run History

View queuing, running, failed, completed, and recovery positions of background tasks; confirm real progress and locate error sources.

Run History (运行记录) on the left is used to view the queuing, running, failed, completed, and recovery positions of background tasks. References to “Task Center” (任务中心) in older docs or prompts point to the same entry. This is a read-only list: it helps you confirm real progress, error reasons, and task sources. The actual operations must be completed in the source workbench.

Confirm Task Status

View whether tasks are queued, running, complete, failed, or waiting to resume.

Failure Diagnosis

Determine whether the issue is model, network, data, or configuration related.

Open Source Page

Run History only shows status; continue, resume, and retry are handled in the workbench.

Common Task Statuses

Run History displays the following common statuses:

Status Change Overview

Status	Meaning	Common Causes
Queued	Task created, waiting for execution resources	Concurrency limit or model busy
Running	Background is calling the model, processing materials, or writing results	Long tasks such as book breakdown, chapter generation
Completed	Task results have been saved	Return to the corresponding module to view outputs
Failed	An error occurred during execution	Model, network, data, or structured output issues
Waiting to Resume	Task chain requires you to confirm the next step or continue from a recovery entry	Pause in Director Follow-up or quality check
Canceled	Task stopped by user or system	Manual cancel or cleanup when switching versions

Run History status does not auto-refresh continuously. When it seems like “nothing is happening,” first refresh the page, then check the associated module to confirm whether results have already been written.

Failure Diagnosis

When a failure occurs, first look at three types of information:

1. Task type: Book start, chapter, book breakdown, knowledge base, or image generation
2. Error source: Model, network, data, structured output, or runtime exception
3. Available results: Whether body text, analysis results, or partial assets were generated

Execution details prioritize showing model service returns for balance, quota, rate-limit, or connection errors. When you see such errors, first handle model connection and balance, then open the source page to continue. Do not treat “chapter step incomplete” as lost body text, and do not re-initiate already completed chapter ranges.

Error Type	Handling Direction
Model or network temporary error	First handle connection or balance, then return to source page to retry
Structured output failure	Go to model settings to adjust model routing, then return to source page to retry
Missing basic data	Return to novel page to supplement information
Available body text but review not perfect	Record quality debt, continue subsequent chapters
Explicit re-planning required	Return to Director Follow-up to handle

Stopping Running Tasks

Running or queued tasks can be stopped directly. After stopping, associated records remain viewable, and saved outputs (novels, chapters, analysis content, source materials) are unaffected.

Director advancement, book breakdown analysis, writing style extraction, cover and character image generation all support directly stopping running tasks. Stopping interrupts model requests currently in waiting, avoiding continued generation of subsequent content or automatic retries.

Deleting Completed Records

Completed (completed, failed, canceled, or waiting to resume) records support confirmation-based deletion one by one. Deleting a run record does not delete saved novels, chapters, analysis content, or source materials; it only cleans up the record list itself.

Queued or running records will prompt you to stop first before deletion, to avoid accidentally deleting records of tasks currently in execution.

How to Distinguish Pause and Failure

Status	Meaning	Recommended Action
Waiting for confirmation / Paused processing	Book rules or checkpoints waiting for user decision	Open source page to view reason and explicitly resume
Failed	Current command did not complete	View real error, then return to source page to decide retry or model switch
Waiting to resume	Background interrupted but retains recoverable position	Open source page, continue from first unfinished step
Completed but with reminder	Output available, still has quality debt or follow-up items	Open source page to verify, do not rerun

Pauses caused by “quality-first” (质量优先) strategy will not be auto-cleared by background polling or Worker. After handling the issue, you need to explicitly resume from Director Follow-up.

Where to Perform Operations

Run History only displays factual status and “open source page.” The actual operations belong in pages that can display creative context, saved outputs, and impact scope:

Task Type	Return to Where to Handle
Auto Director candidate or planning failure	Retry or resume on the AI Auto Director page
Chapter generation or fix failure	Handle on the chapter page
Book breakdown, knowledge base, or image task failure	Handle in the corresponding workbench
Model, balance, or connection issues	Fix in model settings, then return to original task page

Before operating, confirm current task status in Run History. Repeatedly triggering the same stage multiple times may cause status to become difficult to judge.

Auto Director Task Queue Model

Auto Director background actions are written into a command queue, then executed by the director worker. Common status mappings are as follows:

Stage	What the User Sees
Queued	Task queued or waiting to execute
Running	Background is executing; system periodically renews lease to avoid misjudging long tasks
Completed	Task results are viewable
Failed	Run History shows error and source page
Waiting to Resume	Manually resume from source page
Canceled	User or system canceled

This means you can leave the current page and the background will continue running; but it also means repeatedly clicking the same entry may queue multiple commands.

Stale Task Recovery

Stale means the worker lease expired. Common causes include application or service restart, background process exit, long task runtime exceeding lease and renewal failure, local machine sleep, or network interruption. The system distinguishes between auto-recovery and manual recovery:

Failure Recovery Flow

• Full book auto-execution, continue execution, or resume from checkpoint, and attempts not exceeded: auto return to queue, continue execution
• Attempts exceeded or command not suitable for auto-recovery: mark stale, enter waiting to resume
• User clicks resume on source page: continue from most recent checkpoint or command payload

Risk of Repeatedly Starting Same-Scope Tasks

If prompted that an Auto Director task is already processing the same scope, first open Run History to view progress. Repeated starts may make it difficult to judge which task wrote the final result.

Working with Director Follow-up

Recommended troubleshooting order:

1. Confirm Run History

   Confirm whether the current command is still running, and view error details.

2. Determine Status Type

   If waiting for confirmation, enter Director Follow-up to handle the checkpoint; if failed, read the error message first.

3. Open Source Page

   From Run History, open the corresponding workbench to decide retry, resume, or re-plan.

Run History provides facts and source locations; the corresponding workbench provides and executes the next step.

Usage Recommendations

Develop two habits:

• After initiating a long task and encountering a wait, first check Run History
• After a failure, first read the error message, then decide retry, resume, or re-plan from the source page

Run History reduces repeated clicks and blind reruns, serving as a read-only entry point for troubleshooting long chain issues.

Frequently Asked Questions

Why didn't the task status refresh in time?

Run History status requires manually refreshing the page to update. Long-running tasks (such as book breakdown, whole-volume chapter splitting) may last from minutes to hours; refresh to view the latest status.

Will deleting a run record delete my novel?

No. Deletion only clears the entry in the record list; saved novels, chapters, and materials are unaffected.

Can a deleted record be recovered?

Deleted run records cannot be recovered. If the associated task is still executing, stopping and re-triggering will produce a new record.

Run History shows completed, but I don't see content in the novel?

First confirm whether you opened the source page corresponding to that task. Some tasks (such as material organization, material sync) update backend data upon completion and may not display directly in the body text area.

Why are there multiple records for the same task?

Repeatedly clicking the start entry, resume, retry, or re-plan may produce multiple records. When troubleshooting, use the latest record as the reference, and avoid operating on already completed records again.

After a director task pauses, should I click continue or retry?

First view the pause reason. For minor issues or waiting for confirmation, usually choose “continue” to advance from the checkpoint; for explicit failures or model unavailability, first resolve external issues, then “resume” or “retry.”

Director Follow-up

View pause reasons and continue creation in Director Follow-up.

Book Breakdown Workbench

Return to the corresponding workbench after book breakdown, knowledge base, or other task failures.

Model Configuration

Fix model or connection issues in settings first.