pengjunjie/ragflow
1
0
Fork 1
mirror of https://github.com/infiniflow/ragflow.git synced 2026-01-23 03:26:53 +08:00
Code Issues Packages Projects Releases Wiki Activity

Refa: async retrieval process. (#12629)

Browse Source 
### Type of change

- [x] Refactoring
- [x] Performance Improvement
This commit is contained in:
Kevin Hu
2026-01-15 12:28:49 +08:00
committed by GitHub
parent f82628c40c
commit 9a10558f80
11 changed files with 52 additions and 57 deletions
Download Patch File Download Diff File Expand all files Collapse all files

34
api/apps/chunk_app.py
View File

@ -61,7 +61,7 @@ async def list_chunk():
}
if "available_int" in req:
query["available_int"] = int(req["available_int"])
sres = settings.retriever.search(query, search.index_name(tenant_id), kb_ids, highlight=["content_ltks"])
sres = await settings.retriever.search(query, search.index_name(tenant_id), kb_ids, highlight=["content_ltks"])
res = {"total": sres.total, "chunks": [], "doc": doc.to_dict()}
for id in sres.ids:
d = {
@ -371,20 +371,20 @@ async def retrieval_test():
_question += await keyword_extraction(chat_mdl, _question)
labels = label_question(_question, [kb])
ranks = await asyncio.to_thread(settings.retriever.retrieval,
_question,
embd_mdl,
tenant_ids,
kb_ids,
page,
size,
float(req.get("similarity_threshold", 0.0)),
float(req.get("vector_similarity_weight", 0.3)),
doc_ids=local_doc_ids,
top=top,
rerank_mdl=rerank_mdl,
rank_feature=labels,
)
ranks = await settings.retriever.retrieval(
_question,
embd_mdl,
tenant_ids,
kb_ids,
page,
size,
float(req.get("similarity_threshold", 0.0)),
float(req.get("vector_similarity_weight", 0.3)),
doc_ids=local_doc_ids,
top=top,
rerank_mdl=rerank_mdl,
rank_feature=labels
)
if use_kg:
ck = await settings.kg_retriever.retrieval(_question,
@ -413,7 +413,7 @@ async def retrieval_test():
@manager.route('/knowledge_graph', methods=['GET']) # noqa: F821
@login_required
def knowledge_graph():
async def knowledge_graph():
doc_id = request.args["doc_id"]
tenant_id = DocumentService.get_tenant_id(doc_id)
kb_ids = KnowledgebaseService.get_kb_ids(tenant_id)
@ -421,7 +421,7 @@ def knowledge_graph():
"doc_ids": [doc_id],
"knowledge_graph_kwd": ["graph", "mind_map"]
}
sres = settings.retriever.search(req, search.index_name(tenant_id), kb_ids)
sres = await settings.retriever.search(req, search.index_name(tenant_id), kb_ids)
obj = {"graph": {}, "mind_map": {}}
for id in sres.ids[:2]:
ty = sres.field[id]["knowledge_graph_kwd"]

4
api/apps/kb_app.py
View File

@ -373,7 +373,7 @@ async def rename_tags(kb_id):
@manager.route('/<kb_id>/knowledge_graph', methods=['GET']) # noqa: F821
@login_required
def knowledge_graph(kb_id):
async def knowledge_graph(kb_id):
if not KnowledgebaseService.accessible(kb_id, current_user.id):
return get_json_result(
data=False,
@ -389,7 +389,7 @@ def knowledge_graph(kb_id):
obj = {"graph": {}, "mind_map": {}}
if not settings.docStoreConn.index_exist(search.index_name(kb.tenant_id), kb_id):
return get_json_result(data=obj)
sres = settings.retriever.search(req, search.index_name(kb.tenant_id), [kb_id])
sres = await settings.retriever.search(req, search.index_name(kb.tenant_id), [kb_id])
if not len(sres.ids):
return get_json_result(data=obj)

4
api/apps/sdk/dataset.py
View File

@ -481,7 +481,7 @@ def list_datasets(tenant_id):
@manager.route('/datasets/<dataset_id>/knowledge_graph', methods=['GET']) # noqa: F821
@token_required
def knowledge_graph(tenant_id, dataset_id):
async def knowledge_graph(tenant_id, dataset_id):
if not KnowledgebaseService.accessible(dataset_id, tenant_id):
return get_result(
data=False,
@ -497,7 +497,7 @@ def knowledge_graph(tenant_id, dataset_id):
obj = {"graph": {}, "mind_map": {}}
if not settings.docStoreConn.index_exist(search.index_name(kb.tenant_id), dataset_id):
return get_result(data=obj)
sres = settings.retriever.search(req, search.index_name(kb.tenant_id), [dataset_id])
sres = await settings.retriever.search(req, search.index_name(kb.tenant_id), [dataset_id])
if not len(sres.ids):
return get_result(data=obj)

2
api/apps/sdk/dify_retrieval.py
View File

@ -135,7 +135,7 @@ async def retrieval(tenant_id):
doc_ids.extend(meta_filter(metas, convert_conditions(metadata_condition), metadata_condition.get("logic", "and")))
if not doc_ids and metadata_condition:
doc_ids = ["-999"]
ranks = settings.retriever.retrieval(
ranks = await settings.retriever.retrieval(
question,
embd_mdl,
kb.tenant_id,

6
api/apps/sdk/doc.py
View File

@ -935,7 +935,7 @@ async def stop_parsing(tenant_id, dataset_id):
@manager.route("/datasets/<dataset_id>/documents/<document_id>/chunks", methods=["GET"]) # noqa: F821
@token_required
def list_chunks(tenant_id, dataset_id, document_id):
async def list_chunks(tenant_id, dataset_id, document_id):
"""
List chunks of a document.
---
@ -1081,7 +1081,7 @@ def list_chunks(tenant_id, dataset_id, document_id):
_ = Chunk(**final_chunk)
elif settings.docStoreConn.index_exist(search.index_name(tenant_id), dataset_id):
sres = settings.retriever.search(query, search.index_name(tenant_id), [dataset_id], emb_mdl=None, highlight=True)
sres = await settings.retriever.search(query, search.index_name(tenant_id), [dataset_id], emb_mdl=None, highlight=True)
res["total"] = sres.total
for id in sres.ids:
d = {
@ -1559,7 +1559,7 @@ async def retrieval_test(tenant_id):
chat_mdl = LLMBundle(kb.tenant_id, LLMType.CHAT)
question += await keyword_extraction(chat_mdl, question)
ranks = settings.retriever.retrieval(
ranks = await settings.retriever.retrieval(
question,
embd_mdl,
tenant_ids,

2
api/apps/sdk/session.py
View File

@ -1098,7 +1098,7 @@ async def retrieval_test_embedded():
_question += await keyword_extraction(chat_mdl, _question)
labels = label_question(_question, [kb])
ranks = settings.retriever.retrieval(
ranks = await settings.retriever.retrieval(
_question, embd_mdl, tenant_ids, kb_ids, page, size, similarity_threshold, vector_similarity_weight, top,
local_doc_ids, rerank_mdl=rerank_mdl, highlight=req.get("highlight"), rank_feature=labels
)

15
api/db/services/dialog_service.py
View File

@ -403,17 +403,10 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
yield {"answer": msg, "reference": {}, "audio_binary": None, "final": False}
await task
'''
async for think in reasoner.thinking(kbinfos, attachments_ + " ".join(questions)):
if isinstance(think, str):
thought = think
knowledges = [t for t in think.split("\n") if t]
elif stream:
yield think
'''
else:
if embd_mdl:
kbinfos = await asyncio.to_thread(retriever.retrieval,
kbinfos = await retriever.retrieval(
" ".join(questions),
embd_mdl,
tenant_ids,
@ -853,7 +846,7 @@ async def async_ask(question, kb_ids, tenant_id, chat_llm_name=None, search_conf
metas = DocumentService.get_meta_by_kbs(kb_ids)
doc_ids = await apply_meta_data_filter(meta_data_filter, metas, question, chat_mdl, doc_ids)
kbinfos = retriever.retrieval(
kbinfos = await retriever.retrieval(
question=question,
embd_mdl=embd_mdl,
tenant_ids=tenant_ids,
@ -929,7 +922,7 @@ async def gen_mindmap(question, kb_ids, tenant_id, search_config={}):
metas = DocumentService.get_meta_by_kbs(kb_ids)
doc_ids = await apply_meta_data_filter(meta_data_filter, metas, question, chat_mdl, doc_ids)
ranks = settings.retriever.retrieval(
ranks = await settings.retriever.retrieval(
question=question,
embd_mdl=embd_mdl,
tenant_ids=tenant_ids,