OsirisEmbed-v2 — Sovereign Multilingual Embedding Model
768-dimensional multilingual embedding model for OsirisBrain's episodic memory system.
Specs
- Dimensions: 768
- Architecture: nomic-embed-text-v1.5 (Matryoshka)
- Languages: 100+ (English, Spanish, mixed)
- Max tokens: 8,192
- Format: ONNX (quantized)
- Size: ~137 MB
Usage
Used internally by OsirisBrain's embedding_service.ts for:
- Episodic memory vectorization (Hippocampus)
- Semantic memory recall (cosine + temporal decay)
- Conversation auto-memorization
License
Apache 2.0 (inherited from nomic-embed-text-v1.5)