Support resolving AWS Bedrock Application Inference Profiles to base model IDs

[cijaaimee]

Describe the feature or potential improvement

In our company, we utilize AWS Bedrock application inference profiles to track costs granularly per team and/or project.

Because we use these profiles, the traces sent to Langfuse record the inference profile ID rather than the actual underlying base model ID. This creates two significant pain points for us:

• We have to constantly update and maintain complex regex patterns in Langfuse's "Model Definitions" section to ensure all our profile IDs are caught and priced correctly.
• Our usage charts are difficult to read. A single foundational model is represented by many different entries on the dashboard, making it hard to get a high-level view of actual model usage.

We see a few potential ways to solve this:

• Ideally, Langfuse would automatically detect when an AWS Bedrock inference profile ID is being used, resolve it to the underlying base model, and use that base model ID instead. This would completely eliminate the need to clutter our Model Definitions with complex regex patterns.
• Introduce a way to manually group or "alias" multiple IDs to a single canonical model. This would allow us to manually map our various inference profile IDs to their respective base models so they aggregate correctly in the usage charts.
• Instead of displaying the raw model ID ingested from the trace, Langfuse could display the name of the matched Model Definition in the UI and charts. This would at least solve the analytics fragmentation without requiring underlying data structure changes.

Additional information

No response

[CodeCLS]

Thanks for adding this!