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

apps/application/flow/step_node/document_extract_node/i_document_extract_node.py

@@ -23,5 +23,5 @@ class IDocumentExtractNode(INode):
                                                        self.node_params_serializer.data.get('document_list')[1:])
         return self.execute(document=res, **self.flow_params_serializer.data)
 
-    def execute(self, document, **kwargs) -> NodeResult:
+    def execute(self, document, chat_id, **kwargs) -> NodeResult:
         pass

apps/application/flow/step_node/document_extract_node/impl/base_document_extract_node.py

@@ -1,23 +1,64 @@
 # coding=utf-8
 import io
+import mimetypes
 
+from django.core.files.uploadedfile import InMemoryUploadedFile
 from django.db.models import QuerySet
 
 from application.flow.i_step_node import NodeResult
 from application.flow.step_node.document_extract_node.i_document_extract_node import IDocumentExtractNode
 from dataset.models import File
 from dataset.serializers.document_serializers import split_handles, parse_table_handle_list, FileBufferHandle
+from dataset.serializers.file_serializers import FileSerializer
 
 
+def bytes_to_uploaded_file(file_bytes, file_name="file.txt"):
+    content_type, _ = mimetypes.guess_type(file_name)
+    if content_type is None:
+        # 如果未能识别，设置为默认的二进制文件类型
+        content_type = "application/octet-stream"
+    # 创建一个内存中的字节流对象
+    file_stream = io.BytesIO(file_bytes)
+
+    # 获取文件大小
+    file_size = len(file_bytes)
+
+    # 创建 InMemoryUploadedFile 对象
+    uploaded_file = InMemoryUploadedFile(
+        file=file_stream,
+        field_name=None,
+        name=file_name,
+        content_type=content_type,
+        size=file_size,
+        charset=None,
+    )
+    return uploaded_file
+
+
+splitter = '\n`-----------------------------------`\n'
+
 class BaseDocumentExtractNode(IDocumentExtractNode):
-    def execute(self, document, **kwargs):
+    def execute(self, document, chat_id, **kwargs):
         get_buffer = FileBufferHandle().get_buffer
 
         self.context['document_list'] = document
         content = []
-        splitter = '\n`-----------------------------------`\n'
         if document is None or not isinstance(document, list):
-            return NodeResult({'content': content}, {})
+            return NodeResult({'content': ''}, {})
+
+        application = self.workflow_manage.work_flow_post_handler.chat_info.application
+
+        # doc文件中的图片保存
+        def save_image(image_list):
+            for image in image_list:
+                meta = {
+                    'debug': False if application.id else True,
+                    'chat_id': chat_id,
+                    'application_id': str(application.id) if application.id else None,
+                    'file_id': str(image.id)
+                }
+                file = bytes_to_uploaded_file(image.image, image.image_name)
+                FileSerializer(data={'file': file, 'meta': meta}).upload()
 
         for doc in document:
             file = QuerySet(File).filter(id=doc['file_id']).first()
@@ -28,21 +69,21 @@ class BaseDocumentExtractNode(IDocumentExtractNode):
                 if split_handle.support(buffer, get_buffer):
                     # 回到文件头
                     buffer.seek(0)
-                    file_content = split_handle.get_content(buffer)
+                    file_content = split_handle.get_content(buffer, save_image)
                     content.append('## ' + doc['name'] + '\n' + file_content)
                     break
 
         return NodeResult({'content': splitter.join(content)}, {})
 
     def get_details(self, index: int, **kwargs):
+        content = self.context.get('content', '').split(splitter)
         # 不保存content全部内容，因为content内容可能会很大
-        content = (self.context.get('content')[:500] + '...') if len(self.context.get('content')) > 0 else ''
         return {
             'name': self.node.properties.get('stepName'),
             "index": index,
             'run_time': self.context.get('run_time'),
             'type': self.node.type,
-            'content': content,
+            'content': [file_content[:500] for file_content in content],
             'status': self.status,
             'err_message': self.err_message,
             'document_list': self.context.get('document_list')

apps/application/serializers/application_serializers.py

@@ -147,9 +147,9 @@ class ApplicationWorkflowSerializer(serializers.Serializer):
             default_workflow = json.loads(default_workflow_json)
         for node in default_workflow.get('nodes'):
             if node.get('id') == 'base-node':
-                node.get('properties')['node_data'] = {"desc": application.get('desc'),
+                node.get('properties')['node_data']['desc'] = application.get('desc')
-                                                       "name": application.get('name'),
+                node.get('properties')['node_data']['name'] = application.get('name')
-                                                       "prologue": application.get('prologue')}
+                node.get('properties')['node_data']['prologue'] = application.get('prologue')
         return Application(id=uuid.uuid1(),
                            name=application.get('name'),
                            desc=application.get('desc'),
@@ -160,6 +160,14 @@ class ApplicationWorkflowSerializer(serializers.Serializer):
                            model_setting={},
                            problem_optimization=False,
                            type=ApplicationTypeChoices.WORK_FLOW,
+                           stt_model_enable=application.get('stt_model_enable', False),
+                           stt_model_id=application.get('stt_model', None),
+                           tts_model_id=application.get('tts_model', None),
+                           tts_model_enable=application.get('tts_model_enable', False),
+                           tts_model_params_setting=application.get('tts_model_params_setting', {}),
+                           tts_type=application.get('tts_type', None),
+                           file_upload_enable=application.get('file_upload_enable', False),
+                           file_upload_setting=application.get('file_upload_setting', {}),
                            work_flow=default_workflow
                            )
 
@@ -502,6 +510,14 @@ class ApplicationSerializer(serializers.Serializer):
                                type=ApplicationTypeChoices.SIMPLE,
                                model_params_setting=application.get('model_params_setting', {}),
                                problem_optimization_prompt=application.get('problem_optimization_prompt', None),
+                               stt_model_enable=application.get('stt_model_enable', False),
+                               stt_model_id=application.get('stt_model', None),
+                               tts_model_id=application.get('tts_model', None),
+                               tts_model_enable=application.get('tts_model_enable', False),
+                               tts_model_params_setting=application.get('tts_model_params_setting', {}),
+                               tts_type=application.get('tts_type', None),
+                               file_upload_enable=application.get('file_upload_enable', False),
+                               file_upload_setting=application.get('file_upload_setting', {}),
                                work_flow={}
                                )
 

apps/common/handle/base_parse_table_handle.py

@@ -19,5 +19,5 @@ class BaseParseTableHandle(ABC):
         pass
 
     @abstractmethod
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         pass

apps/common/handle/base_split_handle.py

@@ -20,5 +20,5 @@ class BaseSplitHandle(ABC):
         pass
 
     @abstractmethod
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         pass

apps/common/handle/impl/doc_split_handle.py

@@ -190,12 +190,16 @@ class DocSplitHandle(BaseSplitHandle):
             return True
         return False
 
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         try:
             image_list = []
             buffer = file.read()
             doc = Document(io.BytesIO(buffer))
-            return self.to_md(doc, image_list, get_image_id_func())
+            content = self.to_md(doc, image_list, get_image_id_func())
+            if len(image_list) > 0:
+                content = content.replace('/api/image/', '/api/file/')
+                save_image(image_list)
+            return content
         except BaseException as e:
             traceback.print_exception(e)
             return f'{e}'

apps/common/handle/impl/html_split_handle.py

@@ -61,7 +61,7 @@ class HTMLSplitHandle(BaseSplitHandle):
                 'content': split_model.parse(content)
                 }
 
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         buffer = file.read()
 
         try:

apps/common/handle/impl/pdf_split_handle.py

@@ -309,7 +309,7 @@ class PdfSplitHandle(BaseSplitHandle):
             return True
         return False
 
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         with tempfile.NamedTemporaryFile(delete=False) as temp_file:
             # 将上传的文件保存到临时文件中
             temp_file.write(file.read())

apps/common/handle/impl/table/csv_parse_table_handle.py

@@ -35,7 +35,7 @@ class CsvSplitHandle(BaseParseTableHandle):
 
         return [{'name': file.name, 'paragraphs': paragraphs}]
 
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         buffer = file.read()
         try:
             return buffer.decode(detect(buffer)['encoding'])

apps/common/handle/impl/table/xls_parse_table_handle.py

@@ -61,7 +61,7 @@ class XlsSplitHandle(BaseParseTableHandle):
             return [{'name': file.name, 'paragraphs': []}]
         return result
 
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         # 打开 .xls 文件
         try:
             workbook = xlrd.open_workbook(file_contents=file.read(), formatting_info=True)

apps/common/handle/impl/table/xlsx_parse_table_handle.py

@@ -74,7 +74,7 @@ class XlsxSplitHandle(BaseParseTableHandle):
         return result
 
 
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         try:
             # 加载 Excel 文件
             workbook = load_workbook(file)

apps/common/handle/impl/text_split_handle.py

@@ -51,7 +51,7 @@ class TextSplitHandle(BaseSplitHandle):
                 'content': split_model.parse(content)
                 }
 
-    def get_content(self, file):
+    def get_content(self, file, save_image):
         buffer = file.read()
         try:
            return buffer.decode(detect(buffer)['encoding'])

apps/common/job/clean_chat_job.py

@@ -4,6 +4,7 @@ import logging
 import datetime
 
 from django.db import transaction
+from django.db.models.fields.json import KeyTextTransform
 from django.utils import timezone
 from apscheduler.schedulers.background import BackgroundScheduler
 from django_apscheduler.jobstores import DjangoJobStore
@@ -11,6 +12,8 @@ from application.models import Application, Chat
 from django.db.models import Q
 from common.lock.impl.file_lock import FileLock
 from dataset.models import File
+from django.db.models.functions import Cast
+from django.db import models
 
 scheduler = BackgroundScheduler()
 scheduler.add_jobstore(DjangoJobStore(), "default")
@@ -40,7 +43,7 @@ def clean_chat_log_job():
                 break
             deleted_count, _ = Chat.objects.filter(id__in=logs_to_delete).delete()
             # 删除对应的文件
-            File.objects.filter(~Q(meta__chat_id__in=logs_to_delete)).delete()
+            File.objects.filter(meta__chat_id__in=[str(uuid) for uuid in logs_to_delete]).delete()
             if deleted_count < batch_size:
                 break
 

apps/dataset/serializers/file_serializers.py

@@ -61,8 +61,9 @@ class FileSerializer(serializers.Serializer):
     def upload(self, with_valid=True):
         if with_valid:
             self.is_valid(raise_exception=True)
-        file_id = uuid.uuid1()
+        meta = self.data.get('meta')
-        file = File(id=file_id, file_name=self.data.get('file').name, meta=self.data.get('meta'))
+        file_id = meta.get('file_id', uuid.uuid1())
+        file = File(id=file_id, file_name=self.data.get('file').name, meta=meta)
         file.save(self.data.get('file').read())
         return f'/api/file/{file_id}'
 

ui/src/api/dataset.ts

@@ -178,7 +178,7 @@ const putSyncWebDataset: (
 }
 
 /**
- * 重新向量化知识库
+ * 向量化知识库
  * @param 参数 dataset_id
  */
 const putReEmbeddingDataset: (

ui/src/components/ai-chat/ExecutionDetailDialog.vue

@@ -218,26 +218,30 @@
                   <!-- 文档内容提取 -->
                   <template v-if="item.type === WorkflowType.DocumentExtractNode">
                     <div class="card-never border-r-4">
-                      <h5 class="p-8-12">
+                      <h5 class="p-8-12 flex align-center">
-                        参数输出
+                        <span class="mr-4">参数输出</span>
+
                         <el-tooltip
-                            effect="dark"
+                          effect="dark"
-                            content="每个文档仅支持预览500字"
+                          content="每个文档仅支持预览500字"
-                            placement="right"
+                          placement="right"
                         >
                           <AppIcon iconName="app-warning" class="app-warning-icon"></AppIcon>
                         </el-tooltip>
                       </h5>
                       <div class="p-8-12 border-t-dashed lighter">
                         <el-scrollbar height="150">
-                          <MdPreview
+                          <el-card shadow="never" style="--el-card-padding: 8px" v-for="(file_content, index) in item.content"
-                            v-if="item.content"
+                                   :key="index" class="mb-8">
-                            ref="editorRef"
+                            <MdPreview
-                            editorId="preview-only"
+                              v-if="file_content"
-                            :modelValue="item.content"
+                              ref="editorRef"
-                            style="background: none"
+                              editorId="preview-only"
-                          />
+                              :modelValue="file_content"
-                          <template v-else> - </template>
+                              style="background: none"
+                            />
+                            <template v-else> - </template>
+                          </el-card>
                         </el-scrollbar>
                       </div>
                     </div>
@@ -333,6 +337,7 @@
                   <template v-if="item.type === WorkflowType.FormNode">
                     <div class="card-never border-r-4">
                       <h5 class="p-8-12">参数输入</h5>
+
                       <div class="p-8-12 border-t-dashed lighter">
                         <div v-for="(f, i) in item.form_field_list" :key="i" class="mb-8">
                           <span class="color-secondary">{{ f.label.label }}:</span>

ui/src/components/ai-chat/KnowledgeSource.vue

@@ -20,7 +20,12 @@
                 </div>
                 <div class="ml-8" v-else>
                   <a
-                    @click="openLink(item.source_url)"
+                    :href="
+                      item.source_url && !item.source_url.endsWith('/')
+                        ? item.source_url + '/'
+                        : item.source_url
+                    "
+                    target="_blank"
                     class="ellipsis"
                     :title="item?.document_name?.trim()"
                   >
@@ -98,14 +103,6 @@ const uniqueParagraphList = computed(() => {
     }) || []
   )
 })
-
-function openLink(url: string) {
-  // 如果url不是以/结尾，加上/
-  if (url && !url.endsWith('/')) {
-    url += '/'
-  }
-  window.open(url, '_blank')
-}
 </script>
 <style lang="scss" scoped>
 .source_dataset-button {

ui/src/components/ai-chat/component/ParagraphCard.vue

@@ -20,9 +20,25 @@
         <el-text class="flex align-center" style="width: 70%">
           <img :src="getImgUrl(data?.document_name?.trim())" alt="" width="20" class="mr-4" />
 
-          <span class="ellipsis" :title="data?.document_name?.trim()">
+          <template v-if="meta?.source_url">
-            {{ data?.document_name.trim() }}</span
+            <a
-          >
+              :href="
+                meta?.source_url && !meta?.source_url.endsWith('/')
+                  ? meta?.source_url + '/'
+                  : meta?.source_url
+              "
+              target="_blank"
+              class="ellipsis"
+              :title="data?.document_name?.trim()"
+            >
+              {{ data?.document_name?.trim() }}
+            </a>
+          </template>
+          <template v-else>
+            <span class="ellipsis" :title="data?.document_name?.trim()">
+              {{ data?.document_name?.trim() }}
+            </span>
+          </template>
         </el-text>
         <div class="flex align-center" style="line-height: 32px">
           <AppAvatar class="mr-8 avatar-blue" shape="square" :size="18">
@@ -37,6 +53,8 @@
 </template>
 <script setup lang="ts">
 import { getImgUrl } from '@/utils/utils'
+import { computed } from 'vue'
+
 const props = defineProps({
   data: {
     type: Object,
@@ -47,11 +65,22 @@ const props = defineProps({
     default: 0
   }
 })
+const isMetaObject = computed(() => typeof props.data.meta === 'object')
+const parsedMeta = computed(() => {
+  try {
+    return JSON.parse(props.data.meta)
+  } catch (e) {
+    return {}
+  }
+})
+
+const meta = computed(() => (isMetaObject.value ? props.data.meta : parsedMeta.value))
 </script>
 <style lang="scss" scoped>
 .paragraph-source-card-height {
   height: 260px;
 }
+
 @media only screen and (max-width: 768px) {
   .paragraph-source-card-height {
     height: 285px;

ui/src/components/ai-chat/component/operation-button/ChatOperationButton.vue

@@ -19,7 +19,7 @@
       </el-tooltip>
       <el-divider direction="vertical" />
     </span>
-    <span v-if="applicationId && type == 'log'">
+    <span v-if="type == 'ai-chat' || type == 'log'">
       <el-tooltip effect="dark" content="换个答案" placement="top">
         <el-button :disabled="chat_loading" text @click="regeneration">
           <el-icon><RefreshRight /></el-icon>

ui/src/components/ai-chat/component/user-form/index.vue

@@ -7,35 +7,37 @@
     class="mb-16"
     style="padding: 0 24px"
   >
-    <el-card shadow="always" class="dialog-card">
+    <el-card shadow="always" class="dialog-card" style="--el-card-padding: 16px 8px">
-      <div class="flex align-center cursor w-full" @click="showUserInput = !showUserInput">
+      <div class="flex align-center cursor w-full" style="padding: 0 8px;" @click="showUserInput = !showUserInput">
         <el-icon class="mr-8 arrow-icon" :class="showUserInput ? 'rotate-90' : ''"
           ><CaretRight
         /></el-icon>
         用户输入
       </div>
-      <el-collapse-transition>
+      <el-scrollbar max-height="160">
-        <div v-show="showUserInput" class="mt-16">
+        <el-collapse-transition>
-          <DynamicsForm
+          <div v-show="showUserInput" class="mt-16" style="padding: 0 8px;">
-            :key="dynamicsFormRefresh"
+            <DynamicsForm
-            v-model="form_data_context"
+              :key="dynamicsFormRefresh"
-            :model="form_data_context"
+              v-model="form_data_context"
-            label-position="top"
+              :model="form_data_context"
-            require-asterisk-position="right"
+              label-position="top"
-            :render_data="inputFieldList"
+              require-asterisk-position="right"
-            ref="dynamicsFormRef"
+              :render_data="inputFieldList"
-          />
+              ref="dynamicsFormRef"
-          <DynamicsForm
+            />
-            v-if="type === 'debug-ai-chat'"
+            <DynamicsForm
-            v-model="api_form_data_context"
+              v-if="type === 'debug-ai-chat'"
-            :model="api_form_data_context"
+              v-model="api_form_data_context"
-            label-position="top"
+              :model="api_form_data_context"
-            require-asterisk-position="right"
+              label-position="top"
-            :render_data="apiInputFieldList"
+              require-asterisk-position="right"
-            ref="dynamicsFormRef2"
+              :render_data="apiInputFieldList"
-          />
+              ref="dynamicsFormRef2"
-        </div>
+            />
-      </el-collapse-transition>
+          </div>
+        </el-collapse-transition>
+      </el-scrollbar>
     </el-card>
   </div>
 </template>

ui/src/utils/utils.ts

@@ -38,7 +38,7 @@ export function fileType(name: string) {
   获得文件对应图片
 */
 const typeList: any = {
-  txt: ['txt', 'pdf', 'docx', 'csv', 'md', 'html'],
+  txt: ['txt', 'pdf', 'docx', 'md', 'html'],
   table: ['xlsx', 'xls', 'csv'],
   QA: ['xlsx', 'csv', 'xls']
 }

ui/src/views/application-workflow/index.vue

@@ -242,6 +242,9 @@ function clickoutside() {
   showPopover.value = false
 }
 function publicHandle() {
+  // 先执行保存
+  saveApplication()
+  // 后执行发布
   workflowRef.value
     ?.validate()
     .then(() => {

ui/src/views/dataset/DatasetSetting.vue

@@ -148,8 +148,8 @@ async function submit() {
               }
 
         if (cloneModelId.value !== BaseFormRef.value.form.embedding_mode_id) {
-          MsgConfirm(`提示`, `修改知识库向量模型后，需要对知识库重新向量化，是否继续保存？`, {
+          MsgConfirm(`提示`, `修改知识库向量模型后，需要对知识库向量化，是否继续保存？`, {
-            confirmButtonText: '重新向量化',
+            confirmButtonText: '向量化',
             confirmButtonClass: 'primary'
           })
             .then(() => {

ui/src/views/dataset/component/UploadComponent.vue

@@ -129,7 +129,7 @@
         action="#"
         :auto-upload="false"
         :show-file-list="false"
-        accept=".txt, .md, .csv, .log, .docx, .pdf, .html"
+        accept=".txt, .md, .log, .docx, .pdf, .html"
         :limit="50"
         :on-exceed="onExceed"
         :on-change="fileHandleChange"

ui/src/views/dataset/index.vue

@@ -109,7 +109,7 @@
                                 iconName="app-document-refresh"
                                 style="font-size: 16px"
                               ></AppIcon>
-                              重新向量化</el-dropdown-item
+                              向量化</el-dropdown-item
                             >
                             <el-dropdown-item
                               icon="Setting"

ui/src/views/document/index.vue

@@ -23,10 +23,10 @@
               迁移
             </el-button>
             <el-button @click="batchRefresh" :disabled="multipleSelection.length === 0">
-              重新向量化
+              向量化
             </el-button>
             <el-button @click="openGenerateDialog()" :disabled="multipleSelection.length === 0">
-              生成关联问题
+              关联问题
             </el-button>
             <el-button @click="openBatchEditDocument" :disabled="multipleSelection.length === 0">
               设置
@@ -623,7 +623,7 @@ function batchRefresh() {
     }
   })
   documentApi.batchRefresh(id, arr, loading).then(() => {
-    MsgSuccess('批量重新向量化成功')
+    MsgSuccess('批量向量化成功')
     multipleTableRef.value?.clearSelection()
   })
 }
@@ -636,7 +636,7 @@ function batchGenerateRelated() {
     }
   })
   documentApi.batchGenerateRelated(id, arr, loading).then(() => {
-    MsgSuccess('批量生成关联问题成功')
+    MsgSuccess('批量关联问题成功')
     multipleTableRef.value?.clearSelection()
   })
 }

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

@@ -45,7 +45,7 @@
               <img class="mr-12" src="@/assets/icon_file-doc.svg" alt="" />
               <div>
                 <p>文档（TXT、MD、DOCX、HTML、CSV、XLSX、XLS、PDF）</p>
-                <el-text class="color-secondary">需要与文档内容提取节点配合使用</el-text>
+                <el-text class="color-secondary">需要使用“文档内容提取”节点解析文档内容</el-text>
               </div>
             </div>
             <el-checkbox v-model="form_data.document" />
@@ -62,7 +62,7 @@
               <img class="mr-12" src="@/assets/icon_file-image.svg" alt="" />
               <div>
                 <p>图片（JPG、JPEG、PNG、GIF）</p>
-                <el-text class="color-secondary">所选模型需要支持接收图片</el-text>
+                <el-text class="color-secondary">需要使用“图片理解”节点解析图片内容</el-text>
               </div>
             </div>
             <el-checkbox v-model="form_data.image" />

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

@@ -25,7 +25,7 @@
             </div>
             <el-tooltip effect="dark" placement="right" popper-class="max-w-200">
               <template #content>
-                `设置执行该节点输出的内容，{{ '{ from }' }}为表单的占位符。`
+                设置执行该节点输出的内容，{{ '{ form }' }}为表单的占位符。
               </template>
               <AppIcon iconName="app-warning" class="app-warning-icon"></AppIcon>
             </el-tooltip>