[bug]: Embeddings need to be represented in a prompt in human readable, but still load via key

[psychedelicious]

Is there an existing issue for this problem?

• I have searched the existing issues

Operating system

Linux

GPU vendor

Nvidia (CUDA)

GPU model

No response

GPU VRAM

No response

Version number

next

Browser

FF

Python dependencies

No response

What happened

We don't have the concept of embedding triggers - only model keys and names. We need to be able to represent and embedding using a stable "trigger" word (maybe the mutable "name" of the model), and load the embedding in the backend based on this.

Currently on next, embeddings look nice in the prompt, but don't work. This change was in #5801 63f7c61.

Prior to that change, on next embeddings showed up as the long model key in the prompt, but did work.

What you expected to happen

Embeddings are represented as a human-readable string in the prompt and are loaded correctly.

How to reproduce the problem

No response

Additional context

No response

Discord username

No response