Troubleshooting

Record your version, address, product, steps, and full error text first. The checks below prioritize preserving works and candidates.

Symptom	Check first	Next step
Old works missing	Address, port, browser, and profile changes	Return to the original environment and export a backup
AI authentication/permission error	Key, provider, model permissions	Confirm with the provider; avoid repeated resends
Insufficient balance/rate limit	Credit, concurrency, returned limit message	Wait or resolve account limits before retrying
Connection test passes; generation fails	Model capabilities, input length, output limits	Test a small task; retain error text
Browser cross-origin error	Base URL and proxy configuration	Follow settings guidance for a local proxy
Interrupted/empty output	Run state, model response, network	Check for recoverable results first
Stale candidate cannot be adopted	Source changes while waiting	Check the current version before regenerating/editing
Cannot publish	Missing chapters, pending candidates, serious issues, assets	Resolve the checklist item by item
Cannot choose a disk folder	Browser capability, permissions, environment	Reauthorize or use a workspace package
Old JSON rejected	Backup format and supported window	Preserve the original; read compatibility guidance
UI differs from docs	Tagged release, live deployment, current main	Read version guidance and the changelog

Do not start by clearing data

Reloading, clearing cache, and clearing site data are different operations. Clearing site data may delete the local database. Check backups and the original environment first; do not delete manuscript storage to fix a UI issue.

A useful bug report

Use Bug report with version, OS, browser, module, steps, and expected/actual results. Attach screenshots and sanitized logs if useful; never upload API keys or private manuscripts.

If reproduction is unreliable, report that it happened once and describe the conditions rather than guessing the cause.

Continue: Upgrade and compatibility.