Skip to content

3.0.9

Choose a tag to compare

View all tags
@DuckTapeKiller DuckTapeKiller released this 07 Jul 16:40
· 1 commit to main since this release
3.0.9
a6b3420

Horme 3.0.9

LM Studio: automatic model loading

  • Horme now loads your selected LM Studio model automatically before a chat request, instead of relying on LM Studio's unreliable JIT loading. Fixes the "No models loaded" error that occurred when only an embedding model was loaded (e.g. after vault RAG indexing).
  • Uses LM Studio's explicit REST load endpoint (/api/v1/models/load, with a v0 fallback) through Obsidian's requestUrl, so it works despite LM Studio's disabled CORS. Loading the chat model does not evict the embedding model, so RAG indexing keeps working.
  • "Automatic" (empty model) resolves to an already-loaded LLM, or the first non-embedding model available. To control exactly which model loads, set it in Horme's LM Studio settings.
Assets 5
Loading