The OpenAI Plugin enables users to use an OpenAI GPT Model as Machine Translation engine, either for the pretranslation of tasks or with InstantTranslate.
Some Models can be trained within a fine-tuning UI and specific settings for each Model can be made from the Model-properties UI. Generally an OpenAI Model is bound to specific source/target languages and the fine-tuning can only be done for those languages.
After the plugin was activated for the installation some basic configurations have to be made:
OpenAI API URL (runtimeOptions.plugins.translate5AI.OpenAI.server)
This usually points to https://api.openai.com
OpenAI API authorization token (runtimeOptions.plugins.translate5AI.OpenAI.apiToken)
When your OpenAI account is entitled to use their API, one can create API tokens on this page:
https://platform.openai.com/api-keys
https://help.openai.com/en/articles/4936850-where-do-i-find-my-api-key
Make sure the payment-options are setup correctly as it can easily happen requests are rejected due to exhausted token-limits.
Especially "Auto recharge" has to be activated for an uninterrupted use of the API
https://platform.openai.com/account/billing/overview
https://platform.openai.com/account/limits
OpenAI API organization-ID (runtimeOptions.plugins.translate5AI.OpenAI.organizationId, optional)
If you are using multiple OpenAI accounts and/or organisations, this ID can be used to differentiate billing & token limits.
Setup an organization here: https://platform.openai.com/account/organization
OpenAI Machine Translation default matchrate
By default, Machine Pretranslations have a matchrate of 70%. This value can be used to change this setting for OpenAI
Client-level configuration
Available from translate5 version 7.45.0.
The API URL and the API key can be overridden on client level. This way a client can use its own account (and is billed separately), while other clients use the account configured on system level.
| Language resource type | Configs that can be overridden on client level |
|---|---|
| ChatGPT (OpenAI) | runtimeOptions.plugins.translate5AI.OpenAI.server runtimeOptions.plugins.translate5AI.OpenAI.apiToken |
| Azure OpenAI | All runtimeOptions.plugins.translate5AI.Azure.* configs for server, API token and deployment, see Azure |
| translate5 AI | runtimeOptions.plugins.translate5AI.server runtimeOptions.plugins.translate5AI.apiToken |
- Go to Clients, select the client, open the tab System configuration overwrites and set the configs.
- In the language resource overview, click Add language resource and select the resource.
- In Configuration source, select one of the following:
- Use system-level configuration: the language resource uses the system-level configs and can be assigned to several clients in Clients to assign.
- Use client-level configuration: the language resource uses only the configs of the client selected in Client. Only clients that have their own configuration are listed.
- Complete the remaining fields and save.
A language resource with client-level configuration belongs to exactly one client. The configuration source and the client cannot be changed afterwards. If the configs are not set on system level, only Use client-level configuration can be selected.