separate LLM / embedding provider

[jsapede]

is there a way to use ollama for embeddings and openrouter for LLM ?
local LLM with ollama is pain in the ass with small config but embeddings run well, could save some tokens

[InfiniteLoopCoder]

Thank you for your interest in SimpleMem and for the suggestion! Currently, separate providers for the LLM and embedding model are not implemented — the system assumes a unified provider for both. That said, this should be achievable by modifying the LLM layer to route embedding calls and completion calls to different backends independently. If you'd like a quick path forward, modern coding agents can likely handle the refactor with minimal guidance. We think this is a useful feature, so if you get it working, we'd very much welcome a PR. Thanks again!