feat: add collection language config (#1436)

Author: earayu

aperag/api/components/schemas/collection.yaml

@@ -113,11 +113,6 @@ knowledgeGraphConfig:
   type: object
   description: Configuration for knowledge graph generation
   properties:
-    language:
-      type: string
-      description: Language for entity extraction and query responses
-      default: "simplified chinese"
-      example: "simplified chinese"
     entity_types:
       type: array
       items:
@@ -157,8 +152,17 @@ collectionConfig:
       allOf:
         - $ref: '#/knowledgeGraphConfig'
       default:
-        language: "simplified chinese"
         entity_types: ["organization", "person", "geo", "event", "product", "technology", "date", "category"]
+    language:
+      type: string
+      description: Language for the collection content and processing
+      default: "zh-CN"
+      enum:
+        - "zh-CN"
+        - "en-US"
+        - "ja-JP"
+        - "ko-KR"
+      example: "zh-CN"
     embedding:
       $ref: './model.yaml#/modelSpec'
     completion:

aperag/graph/lightrag_manager.py

@@ -83,9 +83,12 @@ async def create_lightrag_instance(collection: Collection) -> LightRAG:
         kg_config = config.knowledge_graph_config
         language = LightRAGConfig.DEFAULT_LANGUAGE
         entity_types = PROMPTS["DEFAULT_ENTITY_TYPES"]
+
+        # Use collection-level language if available
+        if config.language:
+            language = config.language
+
         if kg_config:
-            if kg_config.language:
-                language = kg_config.language
             if kg_config.entity_types:
                 entity_types = kg_config.entity_types
 

aperag/schema/view_models.py

@@ -14,7 +14,7 @@
 
 # generated by datamodel-codegen:
 #   filename:  openapi.merged.yaml
-#   timestamp: 2026-01-29T03:34:01+00:00
+#   timestamp: 2026-02-03T09:01:18+00:00
 
 from __future__ import annotations
 
@@ -71,11 +71,6 @@ class KnowledgeGraphConfig(BaseModel):
     Configuration for knowledge graph generation
     """
 
-    language: Optional[str] = Field(
-        'simplified chinese',
-        description='Language for entity extraction and query responses',
-        examples=['simplified chinese'],
-    )
     entity_types: Optional[list[str]] = Field(
         [
             'organization',
@@ -114,7 +109,6 @@ class CollectionConfig(BaseModel):
     knowledge_graph_config: Optional[KnowledgeGraphConfig] = Field(
         default_factory=lambda: KnowledgeGraphConfig.model_validate(
             {
-                'language': 'simplified chinese',
                 'entity_types': [
                     'organization',
                     'person',
@@ -124,10 +118,15 @@ class CollectionConfig(BaseModel):
                     'technology',
                     'date',
                     'category',
-                ],
+                ]
             }
         )
     )
+    language: Optional[Literal['zh-CN', 'en-US', 'ja-JP', 'ko-KR']] = Field(
+        'zh-CN',
+        description='Language for the collection content and processing',
+        examples=['zh-CN'],
+    )
     embedding: Optional[ModelSpec] = None
     completion: Optional[ModelSpec] = None
     path: Optional[str] = Field(None, description='Path for local and ftp sources')

web/src/api/models/collection-config.ts

@@ -68,6 +68,12 @@ export interface CollectionConfig {
      * @memberof CollectionConfig
      */
     'knowledge_graph_config'?: KnowledgeGraphConfig;
+    /**
+     * Language for the collection content and processing
+     * @type {string}
+     * @memberof CollectionConfig
+     */
+    'language'?: CollectionConfigLanguageEnum;
     /**
      * 
      * @type {ModelSpec}
@@ -201,3 +207,13 @@ export interface CollectionConfig {
     'space_id'?: string;
 }
 
+export const CollectionConfigLanguageEnum = {
+    zh_CN: 'zh-CN',
+    en_US: 'en-US',
+    ja_JP: 'ja-JP',
+    ko_KR: 'ko-KR'
+} as const;
+
+export type CollectionConfigLanguageEnum = typeof CollectionConfigLanguageEnum[keyof typeof CollectionConfigLanguageEnum];
+
+

web/src/api/models/knowledge-graph-config.ts

@@ -20,12 +20,6 @@
  * @interface KnowledgeGraphConfig
  */
 export interface KnowledgeGraphConfig {
-    /**
-     * Language for entity extraction and query responses
-     * @type {string}
-     * @memberof KnowledgeGraphConfig
-     */
-    'language'?: string;
     /**
      * List of entity types to extract during graph indexing
      * @type {Array<string>}

web/src/api/openapi.merged.yaml

@@ -4580,11 +4580,6 @@ components:
       type: object
       description: Configuration for knowledge graph generation
       properties:
-        language:
-          type: string
-          description: Language for entity extraction and query responses
-          default: simplified chinese
-          example: simplified chinese
         entity_types:
           type: array
           items:
@@ -4635,7 +4630,6 @@ components:
           allOf:
             - $ref: '#/components/schemas/knowledgeGraphConfig'
           default:
-            language: simplified chinese
             entity_types:
               - organization
               - person
@@ -4645,6 +4639,16 @@ components:
               - technology
               - date
               - category
+        language:
+          type: string
+          description: Language for the collection content and processing
+          default: zh-CN
+          enum:
+            - zh-CN
+            - en-US
+            - ja-JP
+            - ko-KR
+          example: zh-CN
         embedding:
           $ref: '#/components/schemas/modelSpec'
         completion: