zlei6/ragflow
1
0
Fork 0
mirror of https://github.com/infiniflow/ragflow.git synced 2026-07-12 14:45:42 +08:00
Code Issues Packages Projects Releases Wiki Activity

Handle searching dataset without embedding model (#16742)

Browse Source 
### Summary

Handle searching dataset without embedding model

In this PR, Searching datasets with different embedding models or
searching dataset with/without embedding models are not allowed. We will
improve the behavior later.
This commit is contained in:
qinling0210
2026-07-09 11:38:55 +08:00
committed by GitHub
parent 1430d0e431
commit ae96e636e9
14 changed files with 142 additions and 107 deletions
Download Patch File Download Diff File Expand all files Collapse all files

18
api/db/services/dialog_service.py
View File

@@ -31,7 +31,7 @@ from common.constants import LLMType, ParserType, StatusEnum
from api.db.db_models import DB, Dialog
from api.db.services.common_service import CommonService
from api.db.services.doc_metadata_service import DocMetadataService
from api.db.services.knowledgebase_service import KnowledgebaseService
from api.db.services.knowledgebase_service import KnowledgebaseService, validate_dataset_embedding_models
from api.db.services.langfuse_service import TenantLangfuseService
from api.db.services.llm_service import LLMBundle
from common.metadata_utils import apply_meta_data_filter
@@ -358,16 +358,16 @@ async def async_chat_solo(dialog, messages, stream=True, session_id=None):
def get_models(dialog, trace_context=None, langfuse_session_id=None):
embd_mdl, chat_mdl, rerank_mdl, tts_mdl = None, None, None, None
kbs = KnowledgebaseService.get_by_ids(dialog.kb_ids)
embedding_list = list(set([kb.embd_id for kb in kbs]))
if len(embedding_list) > 1:
raise Exception("**ERROR**: Knowledge bases use different embedding models.")
err = validate_dataset_embedding_models(kbs)
if err:
raise Exception(err)
if embedding_list:
if kbs and kbs[0].embd_id:
embd_owner_tenant_id = kbs[0].tenant_id
embd_model_config = get_model_config_from_provider_instance(embd_owner_tenant_id, LLMType.EMBEDDING, embedding_list[0])
embd_model_config = get_model_config_from_provider_instance(embd_owner_tenant_id, LLMType.EMBEDDING, kbs[0].embd_id)
embd_mdl = LLMBundle(embd_owner_tenant_id, embd_model_config, trace_context=trace_context, langfuse_session_id=langfuse_session_id)
if not embd_mdl:
raise LookupError("Embedding model(%s) not found" % embedding_list[0])
raise LookupError("Embedding model(%s) not found" % kbs[0].embd_id)
if dialog.llm_id:
if dialog.tenant_llm_id:
@@ -721,8 +721,8 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
prompt_config,
partial(
retriever.retrieval,
embd_mdl = embd_mdl,
tenant_ids = tenant_ids,
embd_mdl=embd_mdl,
tenant_ids=tenant_ids,
kb_ids=dialog.kb_ids,
page=1,
page_size=dialog.top_n,

23
api/db/services/knowledgebase_service.py
View File

@@ -29,6 +29,29 @@ from api.constants import DATASET_NAME_LIMIT
from api.utils.api_utils import get_parser_config, get_data_error_result
def _base_model_name(embd_id: str) -> str:
"""Return the base model name by stripping provider/instance suffix from an embd_id."""
parts = embd_id.rsplit("@", 2)
return parts[0]
def validate_dataset_embedding_models(kbs):
"""Validate that all given datasets use the same embedding model (or all use none).
Returns an error message string on failure, or ``None`` on success.
"""
# Either all datasets have an embedding model, or none do. Mixing is not allowed.
embd_ids = [kb.embd_id for kb in kbs if kb.embd_id]
has_embd = len(embd_ids) > 0
if has_embd and len(embd_ids) != len(kbs):
return "Cannot search across datasets where some have embedding models and others do not."
if has_embd:
embd_nms = list({_base_model_name(eid) for eid in embd_ids})
if len(embd_nms) > 1:
return f"Datasets use different embedding models: {[kb.embd_id for kb in kbs]}"
return None
class KnowledgebaseService(CommonService):
"""Service class for managing dataset operations.