Merge branch 'main' of https://github.com/maxkb-dev/maxkb

apps/common/handle/impl/pdf_split_handle.py

@@ -31,6 +31,16 @@ default_pattern_list = [re.compile('(?<=^)# .*|(?<=\\n)# .*'),
 max_kb = logging.getLogger("max_kb")
 
 
+def check_links_in_pdf(doc):
+    for page_number in range(len(doc)):
+        page = doc[page_number]
+        links = page.get_links()
+        if links:
+            for link in links:
+                if link['kind'] == 1:
+                    return True
+    return False
+
 class PdfSplitHandle(BaseSplitHandle):
     def handle(self, file, pattern_list: List, with_filter: bool, limit: int, get_buffer, save_image):
         with tempfile.NamedTemporaryFile(delete=False) as temp_file:
@@ -175,6 +185,9 @@ class PdfSplitHandle(BaseSplitHandle):
 
     @staticmethod
     def handle_links(doc, pattern_list, with_filter, limit):
+        # 检查文档是否包含内部链接
+        if not check_links_in_pdf(doc):
+            return
         # 创建存储章节内容的数组
         chapters = []
         toc_start_page = -1

apps/setting/models_provider/impl/zhipu_model_provider/credential/image.py

@@ -0,0 +1,47 @@
+# coding=utf-8
+import base64
+import os
+from typing import Dict
+
+from langchain_core.messages import HumanMessage
+
+from common import forms
+from common.exception.app_exception import AppApiException
+from common.forms import BaseForm
+from setting.models_provider.base_model_provider import BaseModelCredential, ValidCode
+
+
+class ZhiPuImageModelCredential(BaseForm, BaseModelCredential):
+    api_key = forms.PasswordInputField('API Key', required=True)
+
+    def is_valid(self, model_type: str, model_name, model_credential: Dict[str, object], provider,
+                 raise_exception=False):
+        model_type_list = provider.get_model_type_list()
+        if not any(list(filter(lambda mt: mt.get('value') == model_type, model_type_list))):
+            raise AppApiException(ValidCode.valid_error.value, f'{model_type} 模型类型不支持')
+
+        for key in ['api_key']:
+            if key not in model_credential:
+                if raise_exception:
+                    raise AppApiException(ValidCode.valid_error.value, f'{key} 字段为必填字段')
+                else:
+                    return False
+        try:
+            model = provider.get_model(model_type, model_name, model_credential)
+            res = model.stream([HumanMessage(content=[{"type": "text", "text": "你好"}])])
+            for chunk in res:
+                print(chunk)
+        except Exception as e:
+            if isinstance(e, AppApiException):
+                raise e
+            if raise_exception:
+                raise AppApiException(ValidCode.valid_error.value, f'校验失败,请检查参数是否正确: {str(e)}')
+            else:
+                return False
+        return True
+
+    def encryption_dict(self, model: Dict[str, object]):
+        return {**model, 'api_key': super().encryption(model.get('api_key', ''))}
+
+    def get_model_params_setting_form(self, model_name):
+        pass

apps/setting/models_provider/impl/zhipu_model_provider/model/image.py

@@ -0,0 +1,26 @@
+from typing import Dict
+
+from langchain_openai.chat_models import ChatOpenAI
+
+from common.config.tokenizer_manage_config import TokenizerManage
+from setting.models_provider.base_model_provider import MaxKBBaseModel
+
+
+def custom_get_token_ids(text: str):
+    tokenizer = TokenizerManage.get_tokenizer()
+    return tokenizer.encode(text)
+
+
+class ZhiPuImage(MaxKBBaseModel, ChatOpenAI):
+
+    @staticmethod
+    def new_instance(model_type, model_name, model_credential: Dict[str, object], **model_kwargs):
+        optional_params = MaxKBBaseModel.filter_optional_params(model_kwargs)
+        return ZhiPuImage(
+            model_name=model_name,
+            openai_api_key=model_credential.get('api_key'),
+            openai_api_base='https://open.bigmodel.cn/api/paas/v4',
+            # stream_options={"include_usage": True},
+            streaming=True,
+            **optional_params,
+        )

apps/setting/models_provider/impl/zhipu_model_provider/zhipu_model_provider.py

@@ -11,18 +11,40 @@ import os
 from common.util.file_util import get_file_content
 from setting.models_provider.base_model_provider import ModelProvideInfo, ModelTypeConst, ModelInfo, IModelProvider, \
     ModelInfoManage
+from setting.models_provider.impl.zhipu_model_provider.credential.image import ZhiPuImageModelCredential
 from setting.models_provider.impl.zhipu_model_provider.credential.llm import ZhiPuLLMModelCredential
+from setting.models_provider.impl.zhipu_model_provider.model.image import ZhiPuImage
 from setting.models_provider.impl.zhipu_model_provider.model.llm import ZhipuChatModel
 from smartdoc.conf import PROJECT_DIR
 
 qwen_model_credential = ZhiPuLLMModelCredential()
+zhipu_image_model_credential = ZhiPuImageModelCredential()
+
 model_info_list = [
     ModelInfo('glm-4', '', ModelTypeConst.LLM, qwen_model_credential, ZhipuChatModel),
     ModelInfo('glm-4v', '', ModelTypeConst.LLM, qwen_model_credential, ZhipuChatModel),
     ModelInfo('glm-3-turbo', '', ModelTypeConst.LLM, qwen_model_credential, ZhipuChatModel)
 ]
-model_info_manage = ModelInfoManage.builder().append_model_info_list(model_info_list).append_default_model_info(
+
-    ModelInfo('glm-4', '', ModelTypeConst.LLM, qwen_model_credential, ZhipuChatModel)).build()
+model_info_image_list = [
+    ModelInfo('glm-4v-plus', '具有强大的多模态理解能力。能够同时理解多达五张图像，并支持视频内容理解',
+              ModelTypeConst.IMAGE, zhipu_image_model_credential,
+              ZhiPuImage),
+    ModelInfo('glm-4v', '专注于单图理解。适用于需要高效图像解析的场景',
+              ModelTypeConst.IMAGE, zhipu_image_model_credential,
+              ZhiPuImage),
+    ModelInfo('glm-4v-flash', '专注于单图理解。适用于需要高效图像解析的场景(免费)',
+              ModelTypeConst.IMAGE, zhipu_image_model_credential,
+              ZhiPuImage),
+]
+
+model_info_manage = (
+    ModelInfoManage.builder()
+    .append_model_info_list(model_info_list)
+    .append_default_model_info(ModelInfo('glm-4', '', ModelTypeConst.LLM, qwen_model_credential, ZhipuChatModel))
+    .append_model_info_list(model_info_image_list)
+    .build()
+)
 
 
 class ZhiPuModelProvider(IModelProvider):

installer/Dockerfile

@@ -42,7 +42,7 @@ ENV MAXKB_VERSION="${DOCKER_IMAGE_TAG} (build at ${BUILD_AT}, commit: ${GITHUB_C
     MAXKB_DB_PASSWORD=Password123@postgres \
     MAXKB_EMBEDDING_MODEL_NAME=/opt/maxkb/model/embedding/shibing624_text2vec-base-chinese \
     MAXKB_EMBEDDING_MODEL_PATH=/opt/maxkb/model/embedding \
-    MAXKB_SANDBOX=true \
+    MAXKB_SANDBOX=1 \
     LANG=en_US.UTF-8 \
     PATH=/opt/py3/bin:$PATH \
     POSTGRES_USER=root \

ui/src/components/ai-chat/component/chat-input-operate/index.vue

@@ -6,7 +6,7 @@
         <div
           class="p-8-12"
           v-loading="localLoading"
-          v-if="uploadDocumentList.length || uploadImageList.length"
+          v-if="uploadDocumentList.length || uploadImageList.length || uploadAudioList.length || uploadVideoList.length"
         >
           <el-space wrap>
             <template v-for="(item, index) in uploadDocumentList" :key="index">
@@ -53,6 +53,27 @@
                 />
               </div>
             </template>
+            <template v-for="(item, index) in uploadAudioList" :key="index">
+              <el-card shadow="never" style="--el-card-padding: 8px" class="file cursor">
+                <div
+                  class="flex align-center"
+                  @mouseenter.stop="mouseenter(item)"
+                  @mouseleave.stop="mouseleave()"
+                >
+                  <div
+                    @click="deleteFile(index, 'audio')"
+                    class="delete-icon color-secondary"
+                    v-if="showDelete === item.url"
+                  >
+                    <el-icon><CircleCloseFilled /></el-icon>
+                  </div>
+                  <img :src="getImgUrl(item && item?.name)" alt="" width="24" />
+                  <div class="ml-4 ellipsis" style="max-width: 160px" :title="item && item?.name">
+                    {{ item && item?.name }}
+                  </div>
+                </div>
+              </el-card>
+            </template>
           </el-space>
         </div>
       </el-scrollbar>
@@ -200,7 +221,7 @@ const localLoading = computed({
 const imageExtensions = ['jpg', 'jpeg', 'png', 'gif', 'bmp']
 const documentExtensions = ['pdf', 'docx', 'txt', 'xls', 'xlsx', 'md', 'html', 'csv']
 const videoExtensions = ['mp4', 'avi', 'mov', 'mkv', 'flv']
-const audioExtensions = ['mp3', 'wav', 'aac', 'flac']
+const audioExtensions = ['mp3']
 
 const getAcceptList = () => {
   const { image, document, audio, video } = props.applicationDetails.file_upload_setting
@@ -227,14 +248,14 @@ const getAcceptList = () => {
 const checkMaxFilesLimit = () => {
   return (
     props.applicationDetails.file_upload_setting.maxFiles <=
-    uploadImageList.value.length + uploadDocumentList.value.length
+    uploadImageList.value.length + uploadDocumentList.value.length + uploadAudioList.value.length + uploadVideoList.value.length
   )
 }
 
 const uploadFile = async (file: any, fileList: any) => {
   const { maxFiles, fileLimit } = props.applicationDetails.file_upload_setting
   // 单次上传文件数量限制
-  const file_limit_once = uploadImageList.value.length + uploadDocumentList.value.length
+  const file_limit_once = uploadImageList.value.length + uploadDocumentList.value.length + uploadAudioList.value.length + uploadVideoList.value.length
   if (file_limit_once >= maxFiles) {
     MsgWarning('最多上传' + maxFiles + '个文件')
     fileList.splice(0, fileList.length)
@@ -257,9 +278,9 @@ const uploadFile = async (file: any, fileList: any) => {
   } else if (documentExtensions.includes(extension)) {
     uploadDocumentList.value.push(file)
   } else if (videoExtensions.includes(extension)) {
-    // videos.push(file)
+    uploadVideoList.value.push(file)
   } else if (audioExtensions.includes(extension)) {
-    // audios.push(file)
+    uploadAudioList.value.push(file)
   }
 
 
@@ -297,7 +318,20 @@ const uploadFile = async (file: any, fileList: any) => {
           file.file_id = f[0].file_id
         }
       })
-      console.log(uploadDocumentList.value, uploadImageList.value)
+      uploadAudioList.value.forEach((file: any) => {
+        const f = response.data.filter((f: any) => f.name === file.name)
+        if (f.length > 0) {
+          file.url = f[0].url
+          file.file_id = f[0].file_id
+        }
+      })
+      uploadVideoList.value.forEach((file: any) => {
+        const f = response.data.filter((f: any) => f.name === file.name)
+        if (f.length > 0) {
+          file.url = f[0].url
+          file.file_id = f[0].file_id
+        }
+      })
     })
 }
 const recorderTime = ref(0)
@@ -306,6 +340,8 @@ const recorderLoading = ref(false)
 const inputValue = ref<string>('')
 const uploadImageList = ref<Array<any>>([])
 const uploadDocumentList = ref<Array<any>>([])
+const uploadVideoList = ref<Array<any>>([])
+const uploadAudioList = ref<Array<any>>([])
 const mediaRecorderStatus = ref(true)
 const showDelete = ref('')
 
@@ -433,11 +469,15 @@ function sendChatHandle(event: any) {
       if (inputValue.value.trim()) {
         props.sendMessage(inputValue.value, {
           image_list: uploadImageList.value,
-          document_list: uploadDocumentList.value
+          document_list: uploadDocumentList.value,
+          audio_list: uploadAudioList.value,
+          video_list: uploadVideoList.value,
         })
         inputValue.value = ''
         uploadImageList.value = []
         uploadDocumentList.value = []
+        uploadAudioList.value = []
+        uploadVideoList.value = []
         quickInputRef.value.textareaStyle.height = '45px'
       }
     }
@@ -452,6 +492,10 @@ function deleteFile(index: number, val: string) {
     uploadImageList.value.splice(index, 1)
   } else if (val === 'document') {
     uploadDocumentList.value.splice(index, 1)
+  } else if (val === 'video') {
+    uploadVideoList.value.splice(index, 1)
+  } else if (val === 'audio') {
+    uploadAudioList.value.splice(index, 1)
   }
 }
 function mouseenter(row: any) {

ui/src/components/common-list/index.vue

@@ -73,6 +73,8 @@ defineExpose({
   li {
     padding: 10px 16px;
     font-weight: 400;
+    color: var(--el-text-color-regular);
+    font-size: 14px;
     &.active {
       background: var(--el-color-primary-light-9);
       border-radius: 4px;

ui/src/styles/element-plus.scss

@@ -133,6 +133,7 @@
 
 .el-card {
   --el-card-padding: calc(var(--app-base-px) * 2);
+  color: var(--el-text-color-regular);
 }
 .el-dropdown {
   color: var(--app-text-color);
@@ -267,6 +268,9 @@
 .el-select-group .el-select-dropdown__item {
   padding-left: 11px;
 }
+.el-select-dropdown__item {
+  font-weight: 400;
+}
 
 .el-select__caret {
   color: var(--app-text-color-secondary);

ui/src/views/authentication/component/EditModal.vue

@@ -120,8 +120,14 @@ const open = async (platform: Platform) => {
   let defaultCallbackUrl = window.location.origin
   switch (platform.key) {
     case 'wecom':
+      if (currentPlatform.config.app_key) {
+        currentPlatform.config.agent_id = currentPlatform.config.app_key
+        delete currentPlatform.config.app_key
+      }
+      currentPlatform.config.callback_url = `${defaultCallbackUrl}/api/wecom`
+      break
     case 'dingtalk':
-      if (currentPlatform.config.agent_id && currentPlatform.key === 'dingtalk') {
+      if (currentPlatform.config.agent_id) {
         currentPlatform.config.corp_id = currentPlatform.config.agent_id
         delete currentPlatform.config.agent_id
       }

ui/src/views/template/index.vue

@@ -2,7 +2,7 @@
   <LayoutContainer header="模型设置">
     <div class="template-manage flex main-calc-height">
       <div class="template-manage__left p-8 border-r">
-        <h4 style="padding-bottom: 8px">供应商</h4>
+        <h4 class="p-16">供应商</h4>
         <div class="model-list-height-left">
           <div
             class="all-mode flex cursor"
@@ -33,7 +33,7 @@
                   ref="commonList1"
                 >
                   <template #default="{ row }">
-                    <div class="flex">
+                    <div class="flex align-center">
                       <span
                         :innerHTML="row.icon"
                         alt=""
@@ -59,7 +59,7 @@
                   ref="commonList2"
                 >
                   <template #default="{ row }">
-                    <div class="flex">
+                    <div class="flex align-center">
                       <span
                         :innerHTML="row.icon"
                         alt=""
@@ -301,11 +301,11 @@ onMounted(() => {
   }
 
   .model-list-height {
-    height: calc(var(--create-dataset-height) - 70px);
+    height: calc(var(--create-dataset-height) - 80px);
   }
 
   .model-list-height-left {
-    height: calc(var(--create-dataset-height));
+    height: calc(var(--create-dataset-height) - 40px);
   }
   .all-mode {
     padding: 10px 16px;
@@ -338,6 +338,9 @@ onMounted(() => {
     :deep(.el-collapse-item__wrap) {
       border-bottom: none !important;
     }
+    :deep(.el-collapse-item__content) {
+      padding-bottom: 0 !important;;
+    }
   }
 }
 </style>

ui/src/workflow/nodes/base-node/component/FileUploadSettingDialog.vue

@@ -68,6 +68,23 @@
             <el-checkbox v-model="form_data.image" />
           </div>
         </el-card>
+        <el-card
+          shadow="hover"
+          class="card-checkbox cursor w-full mb-8"
+          :class="form_data.audio ? 'active' : ''"
+          style="--el-card-padding: 8px 16px"
+        >
+          <div class="flex-between">
+            <div class="flex align-center">
+              <img class="mr-12" src="@/assets/icon_file-image.svg" alt="" />
+              <div>
+                <p class="line-height-22 mt-4">音频（MP3）</p>
+                <el-text class="color-secondary">所选模型支持接收音频或与语音转文本节点配合使用</el-text>
+              </div>
+            </div>
+            <el-checkbox v-model="form_data.audio" />
+          </div>
+        </el-card>
       </el-form-item>
     </el-form>
     <template #footer>

ui/src/workflow/nodes/start-node/index.vue

@@ -82,6 +82,12 @@ const refreshFileUploadConfig = () => {
   if (form_data[0].image) {
     fileUploadFields.push({ label: '图片', value: 'image' })
   }
+  if (form_data[0].audio) {
+    fileUploadFields.push({ label: '音频', value: 'audio' })
+  }
+  if (form_data[0].video) {
+    fileUploadFields.push({ label: '视频', value: 'video' })
+  }
 
   set(props.nodeModel.properties.config, 'fields', [...fields, ...fileUploadFields])
 }