runtime:

    environment: prod
    index_namespace: ''
    api_host: 0.0.0.0

    api_port: 6002

    indexer_host: 0.0.0.0

    indexer_port: 6004

    embedding_host: 0.0.0.0

    embedding_port: 6005
    embedding_text_port: 6005
    embedding_image_port: 6008

    translator_host: 0.0.0.0

    translator_port: 6006

    reranker_host: 0.0.0.0

    reranker_port: 6007

  infrastructure:
    elasticsearch:

      host: http://localhost:9200

      username: null
      password: null
    redis:

      host: localhost

      port: 6479
      snapshot_db: 0
      password: null
      socket_timeout: 1
      socket_connect_timeout: 1
      retry_on_timeout: false
      cache_expire_days: 720

      embedding_cache_prefix: embedding
      anchor_cache_prefix: product_anchors

      anchor_cache_expire_days: 30
    database:
      host: null
      port: 3306
      database: null
      username: null
      password: null
    secrets:
      dashscope_api_key: null
      deepl_auth_key: null

  es_index_name: search_products

  indexes: []

  assets:

    query_rewrite_dictionary_path: config/dictionaries/query_rewrite.dict

  product_enrich:
    max_workers: 40

  es_settings:
    number_of_shards: 1
    number_of_replicas: 0

    refresh_interval: 30s

  field_boosts:

    title: 3.0

    qanchors: 2.5
    tags: 2.0
    category_name_text: 2.0
    category_path: 2.0

    brief: 1.5

    description: 1.5

    vendor: 1.5

    option1_values: 1.5
    option2_values: 1.5
    option3_values: 1.5

  query_config:

    supported_languages:

    - zh
    - en
    default_language: en

    enable_text_embedding: true
    enable_query_rewrite: true

    zh_to_en_model: nllb-200-distilled-600m
    en_to_zh_model: nllb-200-distilled-600m
    default_translation_model: nllb-200-distilled-600m
    zh_to_en_model__source_not_in_index: nllb-200-distilled-600m
    en_to_zh_model__source_not_in_index: nllb-200-distilled-600m
    default_translation_model__source_not_in_index: nllb-200-distilled-600m
    translation_embedding_wait_budget_ms_source_in_index: 200
    translation_embedding_wait_budget_ms_source_not_in_index: 300

    style_intent:
      enabled: true

      selected_sku_boost: 1.2

      color_dictionary_path: config/dictionaries/style_intent_color.csv
      size_dictionary_path: config/dictionaries/style_intent_size.csv

      dimension_aliases:

        color:
        - color
        - colors
        - colour
        - colours
        - 颜色
        - 色
        - 色系
        size:
        - size
        - sizes
        - sizing
        - 尺码
        - 尺寸
        - 码数
        - 号码
        - 码

    product_title_exclusion:
      enabled: true

      dictionary_path: config/dictionaries/product_title_exclusion.tsv

    search_fields:
      multilingual_fields:

      - title
      - qanchors
      - category_path
      - category_name_text
      - brief
      - description
      - vendor
      shared_fields: null

      core_multilingual_fields:

      - title
      - qanchors
      - category_name_text

    text_query_strategy:

      base_minimum_should_match: 60%
      translation_minimum_should_match: 60%

      translation_boost: 0.75
      tie_breaker_base_query: 0.5

      best_fields_boost: 2.0
      best_fields:
        title: 4.0
        qanchors: 3.0
        category_name_text: 2.0
      phrase_fields:
        title: 5.0
        qanchors: 4.0
      phrase_match_boost: 3.0

    text_embedding_field: title_embedding
    image_embedding_field: image_embedding.vector

    source_fields:

    - spu_id
    - handle
    - title
    - brief
    - description
    - vendor
    - category_name
    - category_name_text
    - category_path
    - category_id
    - category_level
    - category1_name
    - category2_name
    - category3_name
    - tags
    - min_price
    - compare_at_price
    - image_url
    - sku_prices
    - sku_weights
    - sku_weight_units
    - total_inventory
    - option1_name
    - option1_values
    - option2_name
    - option2_values
    - option3_name
    - option3_values
    - specifications
    - skus

    knn_text_boost: 4
    knn_image_boost: 4

    knn_text_k: 160
    knn_text_num_candidates: 560

    knn_text_k_long: 400
    knn_text_num_candidates_long: 1200

    knn_image_k: 400
    knn_image_num_candidates: 1200

  function_score:

    score_mode: sum
    boost_mode: multiply

    functions: []

  coarse_rank:
    enabled: true
    input_window: 700
    output_window: 240
    fusion:
      text_bias: 0.1
      text_exponent: 0.35

      text_translation_weight: 1.0

      knn_text_weight: 1.0
      knn_image_weight: 1.0
      knn_tie_breaker: 0.1
      knn_bias: 0.6
      knn_exponent: 0.0

  fine_rank:

    enabled: false
    input_window: 160

    output_window: 80
    timeout_sec: 10.0

    rerank_query_template: '{query}'
    rerank_doc_template: '{title}'
    service_profile: fine

  rerank:

    enabled: true

    rerank_window: 160

    timeout_sec: 15.0

    weight_es: 0.4
    weight_ai: 0.6

    rerank_query_template: '{query}'
    rerank_doc_template: '{title}'
    service_profile: default

    fusion:

      rerank_bias: 1.0e-05
      rerank_exponent: 1.15
      fine_bias: 1.0e-05

      fine_exponent: 1.0

      text_bias: 0.1

      text_exponent: 0.25
      text_translation_weight: 0.8

      knn_text_weight: 1.0
      knn_image_weight: 1.0
      knn_tie_breaker: 0.1

      knn_bias: 0.6

      knn_exponent: 0.0

  services:
    translation:

      service_url: http://127.0.0.1:6006
      default_model: nllb-200-distilled-600m
      default_scene: general

      timeout_sec: 10.0

      cache:

        ttl_seconds: 62208000
        sliding_expiration: true

        enable_model_quality_tier_cache: true

        model_quality_tiers:
          deepl: 30
          qwen-mt: 30
          llm: 30
          nllb-200-distilled-600m: 20
          opus-mt-zh-en: 10
          opus-mt-en-zh: 10

      capabilities:

        qwen-mt:

          enabled: true

          backend: qwen_mt
          model: qwen-mt-flash
          base_url: https://dashscope-us.aliyuncs.com/compatible-mode/v1

          timeout_sec: 10.0

          use_cache: true

        llm:

          enabled: true

          backend: llm
          model: qwen-flash
          base_url: https://dashscope-us.aliyuncs.com/compatible-mode/v1

          timeout_sec: 30.0

          use_cache: true

        deepl:

          enabled: true

          backend: deepl
          api_url: https://api.deepl.com/v2/translate

          timeout_sec: 10.0

          glossary_id: ''

          use_cache: true

        nllb-200-distilled-600m:

          enabled: true

          backend: local_nllb
          model_id: facebook/nllb-200-distilled-600M
          model_dir: ./models/translation/facebook/nllb-200-distilled-600M
          ct2_model_dir: ./models/translation/facebook/nllb-200-distilled-600M/ctranslate2-float16
          ct2_compute_type: float16
          ct2_conversion_quantization: float16

          ct2_auto_convert: true

          ct2_inter_threads: 4

          ct2_intra_threads: 0

          ct2_max_queued_batches: 32

          ct2_batch_type: examples
          ct2_decoding_length_mode: source

          ct2_decoding_length_extra: 8
          ct2_decoding_length_min: 32

          device: cuda
          torch_dtype: float16

          batch_size: 64

          max_input_length: 256

          max_new_tokens: 64

          num_beams: 1

          use_cache: true

        opus-mt-zh-en:

          enabled: false

          backend: local_marian
          model_id: Helsinki-NLP/opus-mt-zh-en
          model_dir: ./models/translation/Helsinki-NLP/opus-mt-zh-en
          ct2_model_dir: ./models/translation/Helsinki-NLP/opus-mt-zh-en/ctranslate2-float16
          ct2_compute_type: float16
          ct2_conversion_quantization: float16

          ct2_auto_convert: true
          ct2_inter_threads: 1
          ct2_intra_threads: 0
          ct2_max_queued_batches: 0

          ct2_batch_type: examples
          device: cuda
          torch_dtype: float16

          batch_size: 16
          max_input_length: 256
          max_new_tokens: 256
          num_beams: 1

          use_cache: true

        opus-mt-en-zh:

          enabled: false

          backend: local_marian
          model_id: Helsinki-NLP/opus-mt-en-zh
          model_dir: ./models/translation/Helsinki-NLP/opus-mt-en-zh
          ct2_model_dir: ./models/translation/Helsinki-NLP/opus-mt-en-zh/ctranslate2-float16
          ct2_compute_type: float16
          ct2_conversion_quantization: float16

          ct2_auto_convert: true
          ct2_inter_threads: 1
          ct2_intra_threads: 0
          ct2_max_queued_batches: 0

          ct2_batch_type: examples
          device: cuda
          torch_dtype: float16

          batch_size: 16
          max_input_length: 256
          max_new_tokens: 256
          num_beams: 1

          use_cache: true

    embedding:

      provider: http

      providers:
        http:

          text_base_url: http://127.0.0.1:6005
          image_base_url: http://127.0.0.1:6008
      backend: tei

      backends:
        tei:

          base_url: http://127.0.0.1:8080

          timeout_sec: 20

          model_id: Qwen/Qwen3-Embedding-0.6B

        local_st:

          model_id: Qwen/Qwen3-Embedding-0.6B
          device: cuda

          batch_size: 32
          normalize_embeddings: true

      image_backend: clip_as_service

      image_backends:
        clip_as_service:

          server: grpc://127.0.0.1:51000
          model_name: CN-CLIP/ViT-L-14

          batch_size: 8
          normalize_embeddings: true
        local_cnclip:

          model_name: ViT-L-14

          device: null
          batch_size: 8
          normalize_embeddings: true

    rerank:

      provider: http

      providers:
        http:

          instances:
            default:

              base_url: http://127.0.0.1:6007
              service_url: http://127.0.0.1:6007/rerank

            fine:

              base_url: http://127.0.0.1:6009
              service_url: http://127.0.0.1:6009/rerank

      request:
        max_docs: 1000
        normalize: true

      default_instance: default

      instances:
        default:

          host: 0.0.0.0

          port: 6007

          backend: qwen3_vllm_score
          runtime_dir: ./.runtime/reranker/default

        fine:

          host: 0.0.0.0

          port: 6009

          backend: bge
          runtime_dir: ./.runtime/reranker/fine

      backends:
        bge:

          model_name: BAAI/bge-reranker-v2-m3

          device: null
          use_fp16: true

          batch_size: 80

          max_length: 160

          cache_dir: ./model_cache

          enable_warmup: true

        jina_reranker_v3:

          model_name: jinaai/jina-reranker-v3

          device: null

          dtype: float16

          batch_size: 64

          max_doc_length: 160
          max_query_length: 64
          sort_by_doc_length: true

          cache_dir: ./model_cache

          trust_remote_code: true

        qwen3_vllm:

          model_name: Qwen/Qwen3-Reranker-0.6B
          engine: vllm

          max_model_len: 256

          tensor_parallel_size: 1

          gpu_memory_utilization: 0.2
          dtype: float16

          enable_prefix_caching: true
          enforce_eager: false

          infer_batch_size: 100

          sort_by_doc_length: true

          instruction_format: standard
          instruction: rank products by given query

        qwen3_vllm_score:

          model_name: Qwen/Qwen3-Reranker-0.6B

          use_original_qwen3_hf_overrides: true

          engine: vllm

          max_model_len: 172

          tensor_parallel_size: 1

          gpu_memory_utilization: 0.15

          dtype: float16

          enable_prefix_caching: true
          enforce_eager: false

          infer_batch_size: 80

          sort_by_doc_length: true

          instruction_format: standard
          instruction: Rank products by query with category & style match prioritized

        qwen3_transformers:

          model_name: Qwen/Qwen3-Reranker-0.6B
          instruction: rank products by given query

          max_length: 8192
          batch_size: 64
          use_fp16: true

          attn_implementation: sdpa

        qwen3_transformers_packed:

          model_name: Qwen/Qwen3-Reranker-0.6B
          instruction: Rank products by query with category & style match prioritized

          max_model_len: 256

          max_doc_len: 160
          max_docs_per_pack: 0
          use_fp16: true
          sort_by_doc_length: true

          attn_implementation: eager

        qwen3_gguf:

          repo_id: DevQuasar/Qwen.Qwen3-Reranker-4B-GGUF
          filename: '*Q8_0.gguf'
          cache_dir: ./model_cache
          local_dir: ./models/reranker/qwen3-reranker-4b-gguf
          instruction: Rank products by query with category & style match prioritized

          n_ctx: 512
          n_batch: 512
          n_ubatch: 512
          n_gpu_layers: 999

          main_gpu: 0
          n_threads: 2
          n_threads_batch: 4
          flash_attn: true
          offload_kqv: true
          use_mmap: true
          use_mlock: false
          infer_batch_size: 8
          sort_by_doc_length: true

          length_sort_mode: char

          enable_warmup: true
          verbose: false

        qwen3_gguf_06b:

          repo_id: ggml-org/Qwen3-Reranker-0.6B-Q8_0-GGUF
          filename: qwen3-reranker-0.6b-q8_0.gguf
          cache_dir: ./model_cache
          local_dir: ./models/reranker/qwen3-reranker-0.6b-q8_0-gguf
          instruction: Rank products by query with category & style match prioritized

          n_ctx: 256
          n_batch: 256
          n_ubatch: 256
          n_gpu_layers: 999
          main_gpu: 0
          n_threads: 2
          n_threads_batch: 4
          flash_attn: true
          offload_kqv: true
          use_mmap: true
          use_mlock: false
          infer_batch_size: 32
          sort_by_doc_length: true

          length_sort_mode: char

          reuse_query_state: false
          enable_warmup: true
          verbose: false

        dashscope_rerank:

          model_name: qwen3-rerank
          endpoint: https://dashscope.aliyuncs.com/compatible-api/v1/reranks
          api_key_env: RERANK_DASHSCOPE_API_KEY_CN
          timeout_sec: 10.0
          top_n_cap: 0
          batchsize: 64
          instruct: Given a shopping query, rank product titles by relevance

          max_retries: 2
          retry_backoff_sec: 0.2

  spu_config:
    enabled: true

    spu_field: spu_id

    inner_hits_size: 10

    searchable_option_dimensions:
    - option1
    - option2
    - option3

  tenant_config:

    default:

      primary_language: en
      index_languages:
      - en
      - zh

    tenants:

      '1':
        primary_language: zh
        index_languages:
        - zh
        - en
      '2':
        primary_language: en
        index_languages:
        - en
        - zh
      '3':
        primary_language: zh
        index_languages:
        - zh
        - en
      '162':
        primary_language: zh
        index_languages:
        - zh
        - en
      '170':
        primary_language: en
        index_languages:
        - en
        - zh