Map Model Aliases to Your Chosen LLM Models

This topic describes how to use the Model aliases tab to map LLM model aliases to the real model IDs in your own provider. For more information, see Bring Your Own LLM.

Digital.ai applications call LLM models by a predefined stable alias, such as dai-sage-model, rather than by a specific model ID.

This layer of indirection is what keeps AI features working when models change. Because you can map an alias to several model IDs in priority order, a model that's retired, renamed, or temporarily unavailable at your provider doesn't break the feature: the next model in the list is tried instead.

note

You can map aliases, but you can't create or rename them. Digital.ai maintains the alias list so that applications always have a stable name to call.

Find Your Provider's Model IDs

A mapping holds your provider's own model ID strings, spelled exactly as the provider spells them. Each vendor publishes its current model list, and that's where you copy the value from.

Provider	Where to find the current model IDs
OpenAI	OpenAI models. Open a model to see its Model ID.
Anthropic	Claude models overview. Use the Claude API model ID from the comparison table.
AWS Bedrock	Amazon Bedrock models at a glance. Open a model card to get its exact model ID or inference profile ID, not the friendly name.

warning

Model IDs change over time. Always take the current value from the vendor's list rather than assuming a fixed name. The models available to you also depend on your own account's access and entitlements at the provider. For AWS Bedrock, see Request access to models.

Map an Alias to Your Models

Add at least one provider before you create a mapping. Add mapping stays disabled until your account has one, and the tooltip explains why.

1. Log in to the Platform as an administrator.
2. In the left navigation, click Settings then click AI providers.
3. Click the Model aliases tab, then click Add mapping.
4. In Model alias, select the alias this mapping applies to.
5. In Provider, select the provider this mapping applies to. A mapping always applies to one specific provider.
6. Under Provider models, enter a model ID in the model entry field, or select one of the suggestions, and then click the Add icon. Repeat for each model you want to include.
7. Drag the models by their handles so the model you most prefer is first. Models are tried in the order listed.
8. Click Create.

tip

Model alias and Provider lock after you create the mapping. To change either one, delete the mapping and create it again.

note

If the suggestion list can't load, a warning appears and you can still type model IDs manually. Copy them from your vendor's model list. See Find Your Provider's Model IDs.

Add a Fallback Model or Change Priority

Adding more models to an alias makes it more resilient, because an unavailable model falls through to the next one.

1. On the Model aliases tab, click the Edit icon on the mapping's row.
2. Add models, remove models with the Delete icon, or drag the rows to reorder them.
3. Click Update.

Update a Mapping After a Provider Retires a Model

When your provider retires or renames a model, update the mapping so the alias resolves to a model that still exists.

1. Look up the replacement model ID in your vendor's model list. See Find Your Provider's Model IDs.
2. On the Model aliases tab, click the Edit icon on the mapping's row.
3. Remove the retired model ID and add the replacement.
4. Drag the rows to set the priority order, then click Update.

Delete a Mapping

Deleting a mapping can't be undone. After you delete it, the alias no longer resolves to models in that provider.

1. On the Model aliases tab, click the Delete icon on the mapping's row.
2. In the Delete model mapping dialog, click Delete.

Troubleshoot a Mapping

An AI Feature Isn't Using the Model I Mapped

Work through these checks in order:

1. Confirm the provider connection is valid. See Test a Provider.
2. Confirm the provider named in the mapping is the provider you expect AI operations to use. A mapping applies to one specific provider, and AI operations run through the default provider. See Set the Default Provider.
3. Check that each model ID is spelled exactly as the provider spells it. A single wrong character means the model can't be found.
4. Confirm your provider account has access to that model, in the region you configured. For AWS Bedrock, model access is granted per region.
5. Edit the mapping so it includes a model your provider actually offers, and set the priority order. See Add a Fallback Model or Change Priority.

The Add Mapping Button Is Disabled

You need at least one AI provider before you can create a mapping. See Add a Provider.