jandres/dify
1
0
Fork 0
mirror of https://github.com/langgenius/dify.git synced 2025-12-20 06:32:45 +00:00
Code Issues Actions 5 Packages Projects Releases Wiki Activity

Feat/support multimodal embedding (#29115)

Browse Source 
Co-authored-by: autofix-ci[bot] <114827586+autofix-ci[bot]@users.noreply.github.com>
This commit is contained in:
Jyong
2025-12-09 14:41:46 +08:00
committed by GitHub
parent 77cf8f6c27
commit 9affc546c6
78 changed files with 3230 additions and 713 deletions
Download Patch File Download Diff File Expand all files Collapse all files

25
api/tasks/enable_segment_to_index_task.py
View File

@@ -4,9 +4,10 @@ import time
import click
from celery import shared_task
from core.rag.index_processor.constant.index_type import IndexType
from core.rag.index_processor.constant.doc_type import DocType
from core.rag.index_processor.constant.index_type import IndexStructureType
from core.rag.index_processor.index_processor_factory import IndexProcessorFactory
from core.rag.models.document import ChildDocument, Document
from core.rag.models.document import AttachmentDocument, ChildDocument, Document
from extensions.ext_database import db
from extensions.ext_redis import redis_client
from libs.datetime_utils import naive_utc_now
@@ -67,7 +68,7 @@ def enable_segment_to_index_task(segment_id: str):
return
index_processor = IndexProcessorFactory(dataset_document.doc_form).init_index_processor()
if dataset_document.doc_form == IndexType.PARENT_CHILD_INDEX:
if dataset_document.doc_form == IndexStructureType.PARENT_CHILD_INDEX:
child_chunks = segment.get_child_chunks()
if child_chunks:
child_documents = []
@@ -83,8 +84,24 @@ def enable_segment_to_index_task(segment_id: str):
)
child_documents.append(child_document)
document.children = child_documents
multimodel_documents = []
if dataset.is_multimodal:
for attachment in segment.attachments:
multimodel_documents.append(
AttachmentDocument(
page_content=attachment["name"],
metadata={
"doc_id": attachment["id"],
"doc_hash": "",
"document_id": segment.document_id,
"dataset_id": segment.dataset_id,
"doc_type": DocType.IMAGE,
},
)
)
# save vector index
index_processor.load(dataset, [document])
index_processor.load(dataset, [document], multimodal_documents=multimodel_documents)
end_at = time.perf_counter()
logger.info(click.style(f"Segment enabled to index: {segment.id} latency: {end_at - start_at}", fg="green"))