Module: Legion::Extensions::Llm::Mlx::Runners::Discovery

Extended by:
Discovery
Includes:
Discovery::Pipeline
Included in:
Discovery
Defined in:
lib/legion/extensions/llm/mlx/runners/discovery.rb

Overview

MLX discovery runner: ONLY the MLX-specific work. The generic reconcile / claim / activate / probe (cadence + reactive) / replace / weight-publication / health-display pipeline is mixed in from the shared Discovery::Pipeline. Weight is NOT computed here — the shared WeightReconciler recomputes it from live settings at publish.

The catalog is OpenAI-shaped (GET /v1/models -> body) and readiness is GET /health, so the pipeline's default fetch_raw_models / model_id_from / check_health / health_path are reused. The overrides are the MLX config keys (mlx_api_base / mlx_api_key, defaulting to http://localhost:8000), the Helpers::Callable, and the offering-draft evidence.

Constant Summary collapse

EMBEDDING_PATTERN = 
/embed|bge|e5|nomic/i
UNKNOWN_EVIDENCE_SRC =

Protocol-required evidence source: the default_false taxonomy member.

:default_false

Instance Method Summary collapse

Instance Method Details

#auth_token(instance_cfg:) ⇒ Object

MLX is local: auth is an OPTIONAL bearer key (mlx_api_key, or credentials.api_key when no explicit key is set).



44
45
46
47
 
# File 'lib/legion/extensions/llm/mlx/runners/discovery.rb', line 44

def auth_token(instance_cfg:)
  token = instance_cfg[:mlx_api_key] || instance_cfg.dig(:credentials, :api_key)
  token if token.is_a?(String) && !token.strip.empty?
end

#build_callable(instance_cfg:) ⇒ Object 



49
50
51
 
# File 'lib/legion/extensions/llm/mlx/runners/discovery.rb', line 49

def build_callable(instance_cfg:)
  Legion::Extensions::Llm::Mlx::Helpers::Callable.new(instance_cfg: instance_cfg, logger: log)
end

#build_offering_draft(instance_cfg:, instance_key:, model_id:, model_data:) ⇒ Object

── Offering draft (evidence + metadata; NO weight) ───────────────



54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
 
# File 'lib/legion/extensions/llm/mlx/runners/discovery.rb', line 54

def build_offering_draft(instance_cfg:, instance_key:, model_id:, model_data:)
  tier = instance_cfg[:tier] || :local
  embed_supported = embedding_model?(model_id: model_id)

  Legion::Extensions::Llm::Inventory::OfferingDraft.new(
    provider_native_key: model_id,
    model: model_id,
    tier: tier,
    operation_evidence: build_operation_evidence(embed_supported: embed_supported, model_id: model_id),
    capability_evidence: build_capability_evidence(model_id: model_id),
    context_evidence: build_context_evidence(model_data: model_data),
    max_output_evidence: build_max_output_evidence(model_data: model_data),
    embedding_dimensions_evidence: build_embedding_dimensions_evidence(
      model_data: model_data, embed_supported: embed_supported
    ),
    model_revision_evidence: absent_value_evidence,
    tokenizer_evidence: absent_value_evidence,
    quota_domains: {},
    metadata: build_offering_metadata(model_data: model_data, instance_key: instance_key),
    publication_source: :provider_catalog
  )
end

#catalog_base_url(instance_cfg:) ⇒ Object

── MLX instance-config keys / connection ─────────────────────── Mlx.normalize_instance_config promotes endpoint/base_url/api_base aliases to :mlx_api_base; a missing base falls back to the extension's registered localhost default.



38
39
40
 
# File 'lib/legion/extensions/llm/mlx/runners/discovery.rb', line 38

def catalog_base_url(instance_cfg:)
  normalize_api_base(instance_cfg[:mlx_api_base] || instance_cfg[:endpoint] || 'http://localhost:8000')
end