paraphrase-multilingual-MiniLM-L12-v2
productionfrozen-externaltextinternal
RAG Embedder (MiniLM multilingual).
Frozen multilingual sentence-transformer encoding chunks + queries to 384-dim vectors.
Foundation · paraphrase-multilingual-MiniLM-L12-v2 (Apache 2.0)
What it does
The honest description.
Frozen multilingual sentence-transformer encoding chunks + queries to 384-dim vectors.
Architecture
How it works.
Text chunks (knowledge ingestion) + user queries (retrieval). 384-dim float32 embeddings for cosine similarity in Qdrant.
Training data & drift
What it learned from, and how we watch it.
Source
Frozen upstream weights — no in-house training data path. Replace-in-place via env var to upgrade.
Drift baseline · current quarter
Drift baseline not yet documented for this model.
What it's NOT
- Upstream license: Apache 2.0.
- Swap via RAG_EMBEDDING_MODEL env to upgrade.
Version history · immutable once published
Every release, traceable.
Data Mukodi holds about you is subject to DPDP §12 erasure (use /consent in the app), and any concern about this model can be raised via [email protected] under DPDP §13 grievance, or via /security.