feat: add AI embedding dimension configuration

api/routes/config.py

@@ -1,10 +1,11 @@
 import httpx
 import time
-from fastapi import APIRouter, Depends, Form
+from fastapi import APIRouter, Depends, Form, HTTPException
 from typing import Annotated
 from services.config import ConfigCenter, VERSION
 from services.auth import get_current_active_user, User, has_users
 from api.response import success
+from services.vector_db import VectorDBService
 router = APIRouter(prefix="/api/config", tags=["config"])
 
 
@@ -23,8 +24,27 @@ async def set_config(
     key: str = Form(...),
     value: str = Form(...)
 ):
-    await ConfigCenter.set(key, value)
+    original_value = await ConfigCenter.get(key)
-    return success({"key": key, "value": value})
+    value_to_save = value
+    if key == "AI_EMBED_DIM":
+        try:
+            parsed_value = int(value)
+        except (TypeError, ValueError):
+            raise HTTPException(status_code=400, detail="AI_EMBED_DIM must be an integer")
+        if parsed_value <= 0:
+            raise HTTPException(status_code=400, detail="AI_EMBED_DIM must be greater than zero")
+        value_to_save = str(parsed_value)
+
+    await ConfigCenter.set(key, value_to_save)
+
+    if key == "AI_EMBED_DIM" and str(original_value) != value_to_save:
+        try:
+            service = VectorDBService()
+            service.clear_all_data()
+        except Exception as exc:
+            raise HTTPException(status_code=500, detail=f"Failed to clear vector database: {exc}")
+
+    return success({"key": key, "value": value_to_save})
 
 
 @router.get("/all")

services/processors/vector_index.py

@@ -2,8 +2,9 @@ from typing import Dict, Any
 from fastapi.responses import Response
 import base64
 from services.ai import describe_image_base64, get_text_embedding
-from services.vector_db import VectorDBService
+from services.vector_db import VectorDBService, DEFAULT_VECTOR_DIMENSION
 from services.logging import LogService
+from services.config import ConfigCenter
 
 
 class VectorIndexProcessor:
@@ -71,7 +72,15 @@ class VectorIndexProcessor:
         if embedding is None:
             return Response(content="不支持的文件类型", status_code=400)
 
-        vector_db.ensure_collection(collection_name, vector=True)
+        raw_dim = await ConfigCenter.get('AI_EMBED_DIM', DEFAULT_VECTOR_DIMENSION)
+        try:
+            vector_dim = int(raw_dim)
+        except (TypeError, ValueError):
+            vector_dim = DEFAULT_VECTOR_DIMENSION
+        if vector_dim <= 0:
+            vector_dim = DEFAULT_VECTOR_DIMENSION
+
+        vector_db.ensure_collection(collection_name, vector=True, dim=vector_dim)
         vector_db.upsert_vector(
             collection_name, {'path': path, 'embedding': embedding})
         

services/vector_db.py

@@ -1,6 +1,9 @@
 from pymilvus import CollectionSchema, DataType, FieldSchema, MilvusClient
 
 
+DEFAULT_VECTOR_DIMENSION = 4096
+
+
 class VectorDBService:
     _instance = None
 
@@ -13,15 +16,21 @@ class VectorDBService:
         if not hasattr(self, 'client'):
             self.client = MilvusClient("data/db/milvus.db")
 
-    def ensure_collection(self, collection_name, vector: bool = True):
+    def ensure_collection(self, collection_name, vector: bool = True, dim: int = DEFAULT_VECTOR_DIMENSION):
         if self.client.has_collection(collection_name):
             return
         if vector:
+            try:
+                vector_dim = int(dim)
+            except (TypeError, ValueError):
+                vector_dim = DEFAULT_VECTOR_DIMENSION
+            if vector_dim <= 0:
+                vector_dim = DEFAULT_VECTOR_DIMENSION
             fields = [
                 FieldSchema(name="path", dtype=DataType.VARCHAR,
                             max_length=512, is_primary=True, auto_id=False),
                 FieldSchema(name="embedding",
-                            dtype=DataType.FLOAT_VECTOR, dim=4096)
+                            dtype=DataType.FLOAT_VECTOR, dim=vector_dim)
             ]
             schema = CollectionSchema(
                 fields, description="Image vector collection")

web/src/i18n/locales/en.ts

@@ -202,9 +202,12 @@ export const en = {
   'AI Settings': 'AI Settings',
   'Vision Model': 'Vision Model',
   'Embedding Model': 'Embedding Model',
+  'Embedding Dimension': 'Embedding Dimension',
   'Vector Database': 'Vector Database',
   'Vector Database Settings': 'Vector Database Settings',
   'Database Type': 'Database Type',
+  'Confirm embedding dimension change': 'Confirm embedding dimension change',
+  'Changing the embedding dimension will clear the vector database automatically. You will need to rebuild indexes afterwards. Continue?': 'Changing the embedding dimension will clear the vector database automatically. You will need to rebuild indexes afterwards. Continue?',
   'Confirm clear vector database?': 'Confirm clear vector database?',
   'This will delete all collections irreversibly.': 'This will delete all collections irreversibly.',
   'Confirm Clear': 'Confirm Clear',

web/src/i18n/locales/zh.ts

@@ -204,9 +204,12 @@ export const zh = {
   'AI Settings': 'AI设置',
   'Vision Model': '视觉模型',
   'Embedding Model': '嵌入模型',
+  'Embedding Dimension': '向量维度',
   'Vector Database': '向量数据库',
   'Vector Database Settings': '向量数据库设置',
   'Database Type': '数据库类型',
+  'Confirm embedding dimension change': '确认修改向量维度',
+  'Changing the embedding dimension will clear the vector database automatically. You will need to rebuild indexes afterwards. Continue?': '修改向量维度会自动清空向量数据库，之后需要重建索引，是否继续？',
   'Confirm clear vector database?': '确认清空向量数据库？',
   'This will delete all collections irreversibly.': '此操作将删除所有集合中的所有数据，且不可逆。',
   'Confirm Clear': '确认清空',

web/src/pages/SystemSettingsPage/SystemSettingsPage.tsx

@@ -21,13 +21,16 @@ const VISION_CONFIG_KEYS = [
   { key: 'AI_VISION_API_KEY', label: 'Vision API Key' },
 ];
 
+const DEFAULT_EMBED_DIMENSION = 4096;
+const EMBED_DIM_KEY = 'AI_EMBED_DIM';
+
 const EMBED_CONFIG_KEYS = [
   { key: 'AI_EMBED_API_URL', label: 'Embedding API URL' },
   { key: 'AI_EMBED_MODEL', label: 'Embedding Model', default: 'Qwen/Qwen3-Embedding-8B' },
   { key: 'AI_EMBED_API_KEY', label: 'Embedding API Key' },
 ];
 
-const ALL_AI_KEYS = [...VISION_CONFIG_KEYS, ...EMBED_CONFIG_KEYS];
+const ALL_AI_KEYS = [...VISION_CONFIG_KEYS, ...EMBED_CONFIG_KEYS, { key: EMBED_DIM_KEY, default: DEFAULT_EMBED_DIMENSION }];
 
 // Theme related config keys
 const THEME_KEYS = {
@@ -213,9 +216,27 @@ export default function SystemSettingsPage() {
                 <Form
                   layout="vertical"
                   initialValues={{
-                    ...Object.fromEntries(ALL_AI_KEYS.map(({ key, default: def }) => [key, config[key] ?? def ?? ''])),
+                    ...Object.fromEntries(ALL_AI_KEYS.map(({ key, default: def }) => [key, key === EMBED_DIM_KEY
+                      ? Number(config[key] ?? def ?? DEFAULT_EMBED_DIMENSION)
+                      : config[key] ?? def ?? ''])),
+                  }}
+                  onFinish={async (vals) => {
+                    const currentDim = Number(config[EMBED_DIM_KEY] ?? DEFAULT_EMBED_DIMENSION);
+                    const nextDim = Number(vals[EMBED_DIM_KEY] ?? DEFAULT_EMBED_DIMENSION);
+                    if (currentDim !== nextDim) {
+                      Modal.confirm({
+                        title: t('Confirm embedding dimension change'),
+                        content: t('Changing the embedding dimension will clear the vector database automatically. You will need to rebuild indexes afterwards. Continue?'),
+                        okText: t('Confirm'),
+                        cancelText: t('Cancel'),
+                        onOk: async () => {
+                          await handleSave(vals);
+                        },
+                      });
+                      return;
+                    }
+                    await handleSave(vals);
                   }}
-                  onFinish={handleSave}
                   style={{ marginTop: 24 }}
                   key={JSON.stringify(config)}
                 >
@@ -232,6 +253,9 @@ export default function SystemSettingsPage() {
                         <Input size="large" />
                       </Form.Item>
                     ))}
+                    <Form.Item name={EMBED_DIM_KEY} label={t('Embedding Dimension')}>
+                      <InputNumber min={1} max={32768} style={{ width: '100%' }} />
+                    </Form.Item>
                   </Card>
                   <Form.Item style={{ marginTop: 24 }}>
                     <Button type="primary" htmlType="submit" loading={loading} block>