# Embedding model configuration
embeddings:
  default_model: "all-MiniLM-L6-v2"

  models:
    all-MiniLM-L6-v2:
      provider: sentence-transformers
      dimension: 384
      max_tokens: 256
      description: "Fast, lightweight model for general semantic similarity"

    all-mpnet-base-v2:
      provider: sentence-transformers
      dimension: 768
      max_tokens: 384
      description: "Higher quality embeddings, slower inference"

    text-embedding-3-small:
      provider: openai
      dimension: 1536
      max_tokens: 8191
      description: "OpenAI small embedding model"

    text-embedding-3-large:
      provider: openai
      dimension: 3072
      max_tokens: 8191
      description: "OpenAI large embedding model"

  processing:
    batch_size: 256
    max_workers: 2
    device: "mps"  # cpu, cuda, mps
    normalize: true
    show_progress: true

  storage:
    base_dir: data/embeddings
    format: parquet
    compression: zstd