Schema for Jarvisv2 Embedding Model Settings

---
$id: https://skeleton.botmd.io/cleo/EmbeddingModelSettings.schema
$schema: http://json-schema.org/draft-07/schema#

title: Embedding Model Settings
type: object
description: Defines settings for the embedding model during data and query flow.
properties:
  default_prompt_name:
    default: query
    description: Prompt for the embedding.
    enum: [query, text]
    title: Prompt Name
    type: string
  dimension:
    default: -1
    description: Dimension of the embedding
    title: Dimension
    type: integer
  emb_model_source:
    default: huggingface
    description: Where is the model from?
    enum: [huggingface, bedrock]
    title: Model Source
    type: string
  hnsw_kwargs:
    anyOf:
    - $ref: '#/$defs/HNSWSettings'
    - type: 'null'
    default: null
    description: HNSW settings
    title: HNSW settings
  model:
    description: Name of embedding model to use.
    minLength: 1
    title: Model
    type: string
  query_instruction:
    default: 'Represent this sentence for searching relevant passages:'
    description: Instruction for the query.
    title: Query Instruction
    type: string
  context_size:
    default: 512
    description: Context size of the embedding model
    title: Context Size
    type: integer
required: [model]

$defs:
  HNSWSettings:
    description: |-
      Defines settings for HNSW graph. See
      https://github.com/run-llama/llama_index/blob/977d60a058c691957dae3eb3c66c1894faea24ac/llama-index-integrations/vector_stores/llama-index-vector-stores-postgres/llama_index/vector_stores/postgres/base.py#L570
    properties:
      hnsw_dist_method:
        default: vector_cosine_ops
        description: Distance metric to use. Note that by default PGVectorStore._build_query
          calls cosine_distance
        enum: [vector_l2_ops, vector_ip_ops, vector_cosine_ops, vector_l1_ops, bit_hamming_ops,
          bit_jaccard_ops]
        title: dist_method
        type: string
      hnsw_ef_construction:
        default: 64
        description: Size of the dynamic candidate list for constructing the graph.
          Higher value provides better recall at the cost of speed
        title: ef_construction
        type: integer
      hnsw_ef_search:
        default: 40
        description: Size of the dynamic candidate list for search. Higher value provides
          better recall at the cost of speed.
        title: ef_search
        type: integer
      hnsw_m:
        default: 16
        description: Max number of connections per layer.
        title: m
        type: integer
    title: HNSW Graph Settings
    type: object