Libris Translate 0.28.0 · User guide
Category: Manage Libris Translate
Model providers
Connect the language models your books are translated with.
Steps
- An administrator adds the installation’s providers under Settings › LLM providers. If “Every account may set up its own providers” is allowed, each member can add providers with their own key under My account › My providers.
- Pick the “Connection / protocol”: OpenAI-compatible Chat Completions, OpenAI Chat Completions or Responses with an API key, Anthropic Claude with an API key, or Codex through a ChatGPT account. Enter the address, the model and the key.
- Under “Model and limits” and “Costs and capacity”, set the context window, the maximum output tokens, the timeout, “Concurrent books” and the input and output prices per million tokens that estimates and budgets use.
- Click “Save”, then “Test / detect models”, which checks the connection and lists the models; then choose the provider in the book’s Settings tab.
- For an installation provider, “Shared with the members” decides whether members can see and use it; unticked, it stays reserved to administrators.
What to know
Keys are encrypted on the server and never sent back to the browser, and a saved key is never sent to a new address or connection type until you enter it again. A provider a member adds belongs to that member alone: nobody else sees it or uses its key, and it can only reach public addresses unless the administrator lists private networks in PROVIDER_MEMBER_PRIVATE_NETWORKS. A provider still used by a book or by the request history cannot be deleted; “Permanently retire this provider” stops it while keeping its history. The stronger model and the fallback chain are read from the volume, then its series, then the installation.
Fictional example
Add a fictional local server as “Workshop model” and a hosted API as “Backup model”, share the first with the members, and keep the second for administrators.