fix multi thread app context (#868)

Co-authored-by: jyong <jyong@dify.ai>

api/core/indexing_runner.py

@@ -7,6 +7,7 @@ import time
 import uuid
 from typing import Optional, List, cast
 
+from flask import current_app, Flask
 from flask_login import current_user
 from langchain.schema import Document
 from langchain.text_splitter import RecursiveCharacterTextSplitter, TextSplitter
@@ -522,7 +523,8 @@ class IndexingRunner:
                 sub_documents = all_documents[i:i + 10]
                 for doc in sub_documents:
                     document_format_thread = threading.Thread(target=self.format_qa_document, kwargs={
-                        'tenant_id': tenant_id, 'document_node': doc, 'all_qa_documents': all_qa_documents})
+                        'flask_app': current_app._get_current_object(), 'tenant_id': tenant_id, 'document_node': doc,
+                        'all_qa_documents': all_qa_documents})
                     threads.append(document_format_thread)
                     document_format_thread.start()
                 for thread in threads:
@@ -530,28 +532,29 @@ class IndexingRunner:
             return all_qa_documents
         return all_documents
 
-    def format_qa_document(self, tenant_id: str, document_node, all_qa_documents):
+    def format_qa_document(self, flask_app: Flask, tenant_id: str, document_node, all_qa_documents):
         format_documents = []
         if document_node.page_content is None or not document_node.page_content.strip():
             return
-        try:
-            # qa model document
-            response = LLMGenerator.generate_qa_document(tenant_id, document_node.page_content)
-            document_qa_list = self.format_split_text(response)
-            qa_documents = []
-            for result in document_qa_list:
-                qa_document = Document(page_content=result['question'], metadata=document_node.metadata.copy())
-                doc_id = str(uuid.uuid4())
-                hash = helper.generate_text_hash(result['question'])
-                qa_document.metadata['answer'] = result['answer']
-                qa_document.metadata['doc_id'] = doc_id
-                qa_document.metadata['doc_hash'] = hash
-                qa_documents.append(qa_document)
-            format_documents.extend(qa_documents)
-        except Exception as e:
-            logging.exception(e)
+        with flask_app.app_context():
+            try:
+                # qa model document
+                response = LLMGenerator.generate_qa_document(tenant_id, document_node.page_content)
+                document_qa_list = self.format_split_text(response)
+                qa_documents = []
+                for result in document_qa_list:
+                    qa_document = Document(page_content=result['question'], metadata=document_node.metadata.copy())
+                    doc_id = str(uuid.uuid4())
+                    hash = helper.generate_text_hash(result['question'])
+                    qa_document.metadata['answer'] = result['answer']
+                    qa_document.metadata['doc_id'] = doc_id
+                    qa_document.metadata['doc_hash'] = hash
+                    qa_documents.append(qa_document)
+                format_documents.extend(qa_documents)
+            except Exception as e:
+                logging.exception(e)
 
-        all_qa_documents.extend(format_documents)
+            all_qa_documents.extend(format_documents)
 
 
     def _split_to_documents_for_estimate(self, text_docs: List[Document], splitter: TextSplitter,