Privacy, Cleanup, and Log Sharing

Use this page when you need to free disk space, delete runtime data that contains personal content, or send logs and debug artifacts to someone else for troubleshooting. The per-artifact generation stages and reading order are covered by How to Read and Share a Debug Run; full uninstall by installation type is covered by Uninstall and Data Cleanup; the complete Web admin operations are covered by Administrator Interface.

Scope

• This guide covers: cleaning up logs and runtime data, sanitization rules before external sharing, viewing/exporting desktop and CLI logs and Web session/system logs, and automatic rebuilding of runtime configuration tables.
• This guide does not repeat: where data is stored and sent (see the data-and-privacy page), the meaning of each debug artifact (see the debug-run page), uninstall steps by installation type (see the uninstall page), or every button and state of the Web admin UI (see the administrator-interface page).
• Cleanup is neither uninstall nor backup. Deleting .env, config/, server data/, results, model caches, and debug directories is irreversible; confirm the scope and back up anything you need first.

How to use it

Clean up runtime data

Before cleaning up the desktop app, fully exit the Qt UI (or stop the CLI). Otherwise the log files under result/ are held by the file handler and deletion may fail on Windows:

1. Open “Settings” (Settings) → “General” (General) and read the cleanup hint in the “Verbose Logging” (Verbose Logging) description panel: close the Qt UI first, then delete the unneeded log_*.txt files and the matching timestamp debug folders under result/.
2. Delete log_<timestamp>.txt and the timestamp-image-MD5-size-language-translator/ debug subfolders under result/; delete them together rather than only half of them.
3. manga_translator_work/ next to the input directory may contain per-image JSON, exported text, PSD/JSX, and intermediate images; delete selectively per task.
4. Runtime tables under config/ (filter list, replacements, rich-text rules, translation template, prompts, and so on) are rebuilt with defaults by ensure_runtime_files() on the next start, but custom changes are lost.

The “Cleanup” (Cleanup) module in the Web admin panel only acts on the directories defined by the server cleanup service. It is not an uninstaller that deletes the program directory, and it does not cover the desktop result/ folder or the work directory next to the input directory. “Clear all translation results” on the results page only clears the browser result list and blob URLs; it does not delete result files on the server.

View and export logs

• Desktop and CLI runtime logs are both written to result/log_<yyyyMMddHHmmss>.txt under the application root (the repository root in a dev environment, or the folder of the executable in a packaged build). When the “Translation Error” (Translation Error) dialog appears, click “Open log folder” (Open log folder) to jump to the log directory.
• The “Logs” (Logs) module in the Web admin panel shows system logs and session logs, filtered by session or level, and can “Export Logs” (Export Logs) or “Clear Logs” (Clear Logs).
• Before sharing logs, follow the “Sanitizing and sharing logs” section of this page: remove local paths, credentials, user text, and session tokens, and keep only version, platform, and reproduction steps.

Config import/export and secret display

• The “Export Config” (Export Config) message states that the exported file does not include sensitive information such as API keys; “Import Config” (Import Config) preserves existing sensitive information instead of overwriting it.
• API key fields in the API-management page default to password mode and are shown in the window only when you actively click the reveal icon; this is not protection against clipboard, screen recording, or external services.

Why it happens

Log writing and cleanup boundaries

Run logs are written at DEBUG level to result/log_<yyyyMMddHHmmss>.txt in the same location and format for both the desktop app and the CLI, independent of the console log level. Logs may contain absolute local paths, error stacks, request stages, and session information; they must be sanitized before public sharing.

Server auto cleanup

The server cleanup service cleans only three server data directories:

• results: SERVER_DATA_DIR/results
• user fonts: USER_RESOURCES_DIR/fonts
• user prompts: USER_RESOURCES_DIR/prompts

The settings (admin_settings['cleanup']) default to auto_cleanup: false, interval_hours: 24, max_age_days: 7, and max_size_gb: 10. Each round first deletes files whose modification time is older than the retention period; if the total size still exceeds the limit, it keeps deleting the oldest files until the size is below the limit, then removes empty directories. The service does not cover the desktop result/ folder, manga_translator_work/ next to the input directory, or browser localStorage.

flowchart TD
    A["Cleanup service starts"] --> B{"auto_cleanup enabled?"}
    B -->|no| Z["No cleanup loop starts"]
    B -->|yes| C["Wait interval_hours"]
    C --> D["Delete files older than max_age_days"]
    D --> E{"Total size > max_size_gb?"}
    E -->|yes| F["Delete oldest first until below limit"]
    E -->|no| G["Finish this round"]
    F --> G
    G --> C

Auto cleanup deletes only by modification time and total size; it does not distinguish files that contain sensitive content. Data outside the covered directories must still be handled manually.

Runtime-file rebuild

At startup, the system prepares user-editable runtime tables for every entry point: custom_api_params.json, AI OCR/renderer/colorizer prompts, filter_list.json, text_replacements.yaml, rich_text_rules.yaml, and translation_template.json. It never overwrites user files; a file is deleted and recreated by the follow-up flow only when its content matches the MD5 of a legacy built-in default. Deleting these files restores defaults on the next start, but custom changes are lost, so automatic rebuilding must not be treated as a backup.

Sanitization boundary

• The server recursively masks configuration values (replaced with ***) whose key names contain api_key, api_secret, password, token, or key — and only at that specific output boundary. It is not a general scrubber for logs, debug directories, or databases.
• mask_raw is just a base64-encoded PNG; encoding is not anonymization. PSD/JSX scripts may contain layer text and local file paths; error messages may carry addresses, model names, and request stages. Any file containing this kind of content must be checked item by item before sharing.

Cleanup and sanitization flows

Data cleanup

Data location	Possible contents	Cleanup action	Note
result/log_<timestamp>.txt	Desktop/CLI runtime logs, paths, error stacks	Close the app, then delete	Deletion fails on Windows while the file is held open
result/<timestamp>-<MD5>-<size>-<language>-<translator>/	Verbose debug intermediates, ocrs/, JSON, JSX	Close the app, then delete the whole folder	Delete together with log_*.txt
<input-dir>/manga_translator_work/	Per-image JSON, exported text, PSD/JSX, intermediate images	Delete selectively per task	Contains source images, OCR, translations, and paths
config/ runtime tables	Filter list, replacement/rich-text rules, translation template, prompts	Deleted and rebuilt by ensure_runtime_files() on restart	Custom changes are lost
manga_translator/server/data/	Accounts, sessions, history, logs, user resources	Back up and clean per deployer policy	The cleanup service covers only results/fonts/prompts
Browser localStorage	session_token, result list, user_env_vars	Log out and clear site data	Separate storage from server history

Sanitizing and sharing logs

The goal of sharing logs or debug artifacts is to let the recipient reproduce the issue without your images or keys. Prefer a minimal reproduction set over packaging the whole result/ folder or work directory.

flowchart LR
    A["Prepare logs or debug artifacts to share"] --> B{"Contains sensitive content?"}
    B -->|yes| C["Sanitize file by file: keys/tokens, text, paths, prompts"]
    C --> D["Review again"]
    B -->|no| E["Build a minimal reproduction set"]
    D --> E
    E --> F["Share with version, platform, and reproduction steps"]

Include	Do not include
App/CLI version and operating system	Real API keys, tokens, passwords, session tokens
Reproduction steps, target language, translator, and key parameters	User source images, long source/translation text, OCR text
Sanitized log excerpts and matching debug subfolder	The entire result/ folder or entire work directory
Sanitized config excerpts	Absolute local paths, private prompts

Related settings and limitations

• verbose and the final output directory are independent: debug artifacts are written to result/ under the app root, while final images are written elsewhere per output configuration; cleaning the debug folder does not affect saved outputs.
• The server cleanup service is not uninstall. “Clear all translation results” on the Web results page clears only the browser list and blob URLs, not the result files on the host. See the uninstall page for full uninstall.
• Runtime files are rebuilt with defaults after deletion, but custom content is lost; do not treat automatic rebuilding as a backup or treat example configuration as user configuration.
• Server _sanitize_config applies only to a specific output boundary and does not replace item-by-item review; base64 mask_raw is not sanitization.
• Deleting .env, config/, server data/, results, and model caches is irreversible; logs and error screenshots must be sanitized before external sharing.