estel/dify
1
0
Fork 0
Code Issues Pull Requests Actions 1 Packages Projects Releases Wiki Activity

fix keyword index error when storage source is S3 (#3182)

Browse Source
This commit is contained in:
Jyong
2024-04-09 01:42:58 +08:00
committed by GitHub
parent a81c1ab6ae
commit 283979fc46
2 changed files with 110 additions and 76 deletions
Download Patch File Download Diff File Expand all files Collapse all files

48
api/core/indexing_runner.py
View File

@@ -19,6 +19,7 @@ from core.model_manager import ModelInstance, ModelManager
from core.model_runtime.entities.model_entities import ModelType, PriceType
from core.model_runtime.model_providers.__base.large_language_model import LargeLanguageModel
from core.model_runtime.model_providers.__base.text_embedding_model import TextEmbeddingModel
from core.rag.datasource.keyword.keyword_factory import Keyword
from core.rag.extractor.entity.extract_setting import ExtractSetting
from core.rag.index_processor.index_processor_base import BaseIndexProcessor
from core.rag.index_processor.index_processor_factory import IndexProcessorFactory
@@ -657,18 +658,25 @@ class IndexingRunner:
if embedding_model_instance:
embedding_model_type_instance = embedding_model_instance.model_type_instance
embedding_model_type_instance = cast(TextEmbeddingModel, embedding_model_type_instance)
with concurrent.futures.ThreadPoolExecutor(max_workers=10) as executor:
futures = []
for i in range(0, len(documents), chunk_size):
chunk_documents = documents[i:i + chunk_size]
futures.append(executor.submit(self._process_chunk, current_app._get_current_object(), index_processor,
chunk_documents, dataset,
dataset_document, embedding_model_instance,
embedding_model_type_instance))
# create keyword index
create_keyword_thread = threading.Thread(target=self._process_keyword_index,
args=(current_app._get_current_object(),
dataset, dataset_document, documents))
create_keyword_thread.start()
if dataset.indexing_technique == 'high_quality':
with concurrent.futures.ThreadPoolExecutor(max_workers=10) as executor:
futures = []
for i in range(0, len(documents), chunk_size):
chunk_documents = documents[i:i + chunk_size]
futures.append(executor.submit(self._process_chunk, current_app._get_current_object(), index_processor,
chunk_documents, dataset,
dataset_document, embedding_model_instance,
embedding_model_type_instance))
for future in futures:
tokens += future.result()
for future in futures:
tokens += future.result()
create_keyword_thread.join()
indexing_end_at = time.perf_counter()
# update document status to completed
@@ -682,6 +690,24 @@ class IndexingRunner:
}
)
def _process_keyword_index(self, flask_app, dataset, dataset_document, documents):
with flask_app.app_context():
keyword = Keyword(dataset)
keyword.create(documents)
if dataset.indexing_technique != 'high_quality':
document_ids = [document.metadata['doc_id'] for document in documents]
db.session.query(DocumentSegment).filter(
DocumentSegment.document_id == dataset_document.id,
DocumentSegment.index_node_id.in_(document_ids),
DocumentSegment.status == "indexing"
).update({
DocumentSegment.status: "completed",
DocumentSegment.enabled: True,
DocumentSegment.completed_at: datetime.datetime.utcnow()
})
db.session.commit()
def _process_chunk(self, flask_app, index_processor, chunk_documents, dataset, dataset_document,
embedding_model_instance, embedding_model_type_instance):
with flask_app.app_context():
@@ -700,7 +726,7 @@ class IndexingRunner:
)
# load index
index_processor.load(dataset, chunk_documents)
index_processor.load(dataset, chunk_documents, with_keywords=False)
document_ids = [document.metadata['doc_id'] for document in chunk_documents]
db.session.query(DocumentSegment).filter(