Different chat and embedding models #865
Description
Describe the bug
I verified searxng was working with json. I'm getting an error in the logs "app-1 | error: 'this model does not support embeddings',
app-1 | status_code: 500,
app-1 | attemptNumber: 7,"
Seems that Perplexica is calling my main inference engine and not the embedding engine. I had this configured and working a few revisions ago (before Perplexica got the userid/password added). But now the logs show it calling the main inference model for embeddings. Is this something I configured wrong (but I don't think my config changed...)