Latest
v0.9.4
Adds custom embedding models for semantic search and improves model switching, browser sign-in recovery, credential storage, and panel startup.
What's New
- Custom embedding models: Use your own OpenAI, Google, OpenAI-compatible, OpenRouter, or Ollama embedding model for semantic search. Configure the provider and model in Settings, with output dimensions detected automatically when testing or saving a new model.
Improvements & Fixes
- Embedding model settings: Clearer model configuration and connection errors. Switching embedding models asks for confirmation before rebuilding the semantic index; your original library items are not affected.
- Browser sign-in recovery: A dedicated authorization screen provides options to reopen the browser, copy the sign-in link, or enter an authorization code for browser login. Failed manual attempts keep your input available for retry.
- Credential storage compatibility: Improved saved-credential updates across Zotero's underlying browser engine versions.
- Panel startup: Improved detection of the frontend becoming ready when opening BibGenie.