Introduction

To access TMS, MT, or LLM translation connection, navigate to the Settings page using the cog icon:

View Translation Connections

[Administrator, Manager, Developer, Translator]

All users can view the Translation Connections, but only the administrator can create and modify them. They can only be modified when the system is in maintenance mode.

For systems with large number of Translation Connections, filtering, searching and sorting is available. Furthermore, clicking on a Translation Connection card shows details for that connection.

Create and Configure Connections

[Administrator]

The administrator user can select the (Create new MT connection) button on the Manage MT Connections page to add new connections. To create a machine translator, the first step is to determine which translator will be connected to. Then provide the system with the appropriate information for that translator.

Large Language Models (LLM)

LLMs are the up and coming translation systems. When connecting to an LLM, pick the model that fits your needs. LLMs may be public or private, that is the data sent to the LLM may be used outside the company or be kept solely within the company.

Each LLM type supports multiple models. For instance, Mistral is an LLM that supports over 20 models at the time of this writing, such as mistral-medium-3.5 or mistral-nemo-12B, each with its own specialized intent. Each LLM provider updates their model often.

Some LLM models support passing image as context, some don't. If your projects are configured with Localyzer InContext, make sure the LLM model will make use of it. For example, Claude model claude-haiku-4-5 does not seem to pay much attention to an InContext image passed for translation, while Claude model claude-sonnet-4.6 does.

Some are better suited for translation than others.

To create an LLM connection, please see: Create LLM Connection

LLM using OpenAI

https://api.openai.com/v1/chat/completions

Localyzer uses the OpenAI API to connect to many LLMs, such as, but not limited, to:

ChatGPT (e.g.: gpt-5.4-mini)
Mistral (e.g.: mistral-large-latest, mistral-medium-3-5. Note: might be best suited for European based companies)
Grok (e.g.: grok-beta)
etc.

Other LLM Types

The following require a different connection method than the OpenAI API:

Claude : https://api.anthropic.com/v1/messages -- e.g. claude-haiku-4-5, claude-sonnet-4-6
Gemini: https://generativelanguage.googleapis.com/v1beta/models/ -- e.g. gemini-3.5-flash

Machine Translation (MT)

Translation Management Systems (TMS)

(Obsolete: Systran)

Edit Translation Connections

Connections can only be edited when the system is in Maintenance Mode which will be indicated at the top of the screen. The administrator user can hover over the credential to change and the Edit button will appear. If the system is not in maintenance mode, the lock icon will appear.

Delete Translation Connections

When in maintenance mode, select the connections from the list to be deleted and the delete icon will appear at the top of the screen. Select the icon.

Assign Translation Connections

The MT connection is assigned to a project when created or modified.

Click on a card to configure that TMS connection. Typical configuration requirements include Identifiers for the specific instance or account of a TMS.

By default, a list of locale codes are provided. The display names should remain consistent: For instance, French (France) should be used as a display name. The corresponding code may be fr-FR or fr-France or just fr, depending on the TMS. This will allow an easy switch from one TMS or MT to another if necessary.

If a TMS sends a callback to inform Localyzer of a translation, an API User must first be configured in the Users panel and assigned to that TMS.

For more information, check the links below:

InContext Server and Translation Connections

[Administrator]

Context, captured from Figma, Screenshots, or Google Chrome extension for Webapps, can be passed to a TMS via a PXML file. An InContext Server must first be associated with a TMS.

Enable the 'Convert to PXML' checkbox and fill in the InContext server information for best result to pass context.

Prompting at the Translation Connection Level

Prompting is described here: Prompting

Prompting can be passed to

a TMS - e.g. Trados Enterprise - using PXML
an LLM connection - e.g. ChatGPT

The prompt that is passed to either is a combination of :

The Translation Connection Level prompt
The Project Level Prompt
If InContext Server has a string associated image, that image for that string
The text to translation

The Translation Connection Level prompt is set here:

Note: The variables ${SOURCE_LOCALE} and ${TARGET_LOCALE} must be part of the prompt.

Search and Sort Connections