chat:
  default_model: primary
  backend:
    type: direct_sdk
  timeouts:
    request_seconds: 60
  retry:
    same_deployment_attempts: 2
    max_deployment_failovers: 1
    max_model_fallbacks: 1
  security:
    allowed_providers: [openai]
  models:
    primary:
      deployments:
        - name: openai-primary
          provider: ${HARBOR_CHAT_PROVIDER}
          model: ${HARBOR_CHAT_MODEL}
          api_key: ${HARBOR_CHAT_API_KEY}
          capabilities:
            streaming: true
            structured_output: true
            json_mode: true
            tools: true

embed:
  default_model: primary
  timeouts:
    request_seconds: 60
  default_batch_size: 64
  security:
    allowed_providers: [openai, ollama, huggingface]
  models:
    primary:
      embedding_space: harbor-production-v1
      deployments:
        - name: openai-embedding
          provider: ${HARBOR_EMBED_PROVIDER}
          model: ${HARBOR_EMBED_MODEL}
          api_key: ${HARBOR_EMBED_API_KEY}
          expected_dimensions: 1536
          capabilities:
            batch: true
            configurable_dimensions: true
            default_dimensions: 1536
            encoding_format: true
            purpose: true
            supported_purposes: [query, document]