Privacy and Your Data

Polygluttony has no telemetry, no analytics, no crash reporting and no update checks. It talks to the AI providers you connect, and nothing else.

Who Polygluttony talks to

Server	When
api.anthropic.com	Only if you connected an Anthropic model
api.openai.com	Only if you connected an OpenAI model
generativelanguage.googleapis.com	Only if you connected a Google model
localhost:11434 / localhost:1234	Ollama / LM Studio, on your own computer
The address you entered	Only for an Other model

That's every request it makes: testing a model (one tiny message), loading the model list, the glossary steps and translating. The Get a key at… links open in your browser, and only when you click them. The desktop notification at the end of a run comes from your own system.

Look up names online happens on the provider's side, using its own search tool. Polygluttony doesn't visit any website itself.

What gets sent

Step	Sent to the model
Translate	The prompt, your tone, the dialogue text of each batch, the glossary terms in that batch, and up to 7 earlier lines with their translations for context. Retranslate line in an episode window works the same way.
Verify	A sample of source lines with their translations
Find names	The dialogue text of the episodes being read, plus any terms from your translated episodes
Tidy up	Glossary terms, one category at a time: the new ones in a build, or the ones you picked under Improve
Look up names online	The show name, and the glossary terms it should check
Import from translated episodes	The dialogue of the translated files you picked

Never sent: file names, folder paths, styles, timings, fonts or anything else from the file's header. Formatting tags are removed from the text before it goes out.

What happens to your text on the provider's side is up to their data policy. Check it, especially for unreleased material. A local model through Ollama or LM Studio keeps everything on your computer.

Your API keys

Keys are stored in config.json in Polygluttony's data folder (below), as plain text, not in your system's keychain. They're never sent anywhere except to their own provider. Anyone who can read your user files can read them too, so treat that folder like you'd treat a password file.

There's one saved key per cloud provider. Removing the last model from a provider deletes its key.

Two more files hold your keys, also as plain text:

• Models → Export setup… writes them into the export file. That file never leaves your computer unless you move it. See Moving to another computer.
• Models → Import setup… first saves the setup it replaces as setup-backup-*.zip in the data folder, so you can undo the import. The newest 3 are kept.

Where your data lives

OS	Polygluttony's data folder
Windows	%APPDATA%\uk.blyat.polygluttony
macOS	~/Library/Application Support/uk.blyat.polygluttony
Linux	~/.local/share/uk.blyat.polygluttony (or $XDG_DATA_HOME/uk.blyat.polygluttony)

Inside it:

File	Holds
config.json	Your models, API keys, per-step settings and default languages
projects.json	Recent folders, and each folder's languages, world type, tone and whether its glossary tick is on
results/	Each episode's last result: its status, open issues and translated lines. This is what Review and Resume use, so they work after a restart.
prompts/	Prompts you changed on the Prompts page
setup-backup-*.zip	Only after an Import setup…: the setup it replaced, API keys included
config.v1.backup.json	Only if you upgraded from 1.x: your old settings, kept once before converting them

results/ holds copies of your dialogue and its translations. It stays on your computer like everything else here.

There are no log files. The Logs panel lives in memory and is gone when you close the app.

In your subtitle folders

Translations, glossary.json, glossary-reference.json, and a hidden .polygluttony folder with the glossary's notes and its last 10 versions. A folder used with 2.x may also have glossary.prev.json. Uninstalling leaves them where they are. See Folders & files.