| description | Base model settings, model mutability, and legal requirements such as processing zone and document storage. |
|---|---|
| icon | sliders-up |
On the Model Settings page, you'll have the opportunity to modify the model name and the cover image.
Feel free to set them up so that the Model will be easy to find among all your other Models.
Every model has a unique ID which is generated at creation time.
{% hint style="info" %} Model Templates in the Catalog do not have a Model ID.
The Model ID is generated when the Model is created from the Template. If two users create a Model from the same Template, each Model gets its own unique Model ID. {% endhint %}
You will need the Model ID in order to use the Mindee API.
In the Model Settings page, you can view and copy the Model ID.
The Mindee Support team may request the Model ID to diagnose issues and provide solutions.
{% include "../.gitbook/includes/model-processing-zone.md" %}
{% include "../.gitbook/includes/model-storage-policy.md" %}
It can be useful to copy an existing model for some types of workflows.
You can have a base "template" model that is not called directly, but is used to make derivative models. This way you can have a common base and then modify the Data Schema to account for different providers, geographies, downstream users, etc. Each of these derivative models would be a copy of the "template" model.
You can also us this as way for testing changes to a model. For example you can copy a model used in production, modify the copy, and test the modifications in staging. Once the modifications are tested successfully, switch production over to the new model.
Each copied model has its own ID, to switch models in this way simply set the new/copied model's ID in your integration.
By default, copying the model also copies:
- the model's optional features configuration
- any RAG documents in the model's database
If you do not wish to copy these, simply uncheck the relevant checkboxes in the "Copy Model" dialog box.
All model copies stay within the same organization, ensuring protection of sensitive data.
To prevent unintended changes once your model configuration is finalized, you have the option to lock the model.
This will prevent all changes to settings that change the output of the model. These include the model configuration (Data Schema, class definitions, chaining, ...) and some optional features like confidence scores.
Data processing options are not impacted by the lock.
This is useful for ensuring that the responses given by the model remain stable for production use.
{% hint style="warning" %} Locking the Model is an irreversible action, you will not be able to unlock it afterwards.
Note: you can copy a locked model and modify the copy. {% endhint %}
Models are always active and fully usable.
There is no limit to the number of models you can have.
If you no longer have use of a model, you can delete it at any time.
{% hint style="danger" %} Deleting a model is an irreversible action, all data will be lost forever! {% endhint %}
Deleting a model also deletes:
- Documents and results stored in the Documents History
- Documents and annotations in the RAG database
- All Insights, including API calls