Arivu 1.8.4
Embeddings now work with providers other than Gemini, with their own model setting and a safe background rebuild when you switch.
Arivu 1.8.4 is out. It fixes a gap that affected anyone running Arivu with a provider other than Gemini.
What’s in it
- Embeddings work with your provider. Semantic ranking in Graph search, related items, Graph similarity links, and similar-content duplicates used to work only with Gemini. They now work with OpenAI, OpenRouter, Mistral, Fireworks AI, Perplexity, Ollama, LM Studio, and any compatible endpoint. Read Semantic connections now work with your provider.
- A separate embedding model setting. Admin settings now have a Text generation model group and an Embedding model group. The embedding model can follow your text provider, use a different one, or be turned off.
- Clearer messages. When embeddings are not set up, Arivu says so plainly and lists the features affected, instead of reporting a text provider problem.
- Safe switching. Changing the embedding model rebuilds older embeddings in the background, and a wrong model name pauses that work with a visible error.
- Fixed provider presets. LM Studio and Hugging Face were saved as Custom. They now keep their own preset.
Before you upgrade
On Gemini, nothing changes. On another provider that offers embeddings, Arivu starts embedding your existing library in the background on first start, which uses that provider’s quota. Choose a different embedding model, or Off, in Admin settings if you do not want that.
Follow Self-Hosting Arivu for upgrade steps.