Feat: add meta data filter. (#9405)

### What problem does this PR solve?

#8531 
#7417 
#6761 
#6573
#6477

### Type of change

- [x] New Feature (non-breaking change which adds functionality)
This commit is contained in:
Kevin Hu
2025-08-12 14:12:56 +08:00
committed by GitHub
parent 3ccaa06031
commit 153e430b00
11 changed files with 184 additions and 4 deletions

View File

@ -51,6 +51,7 @@ def set_dialog():
similarity_threshold = req.get("similarity_threshold", 0.1)
vector_similarity_weight = req.get("vector_similarity_weight", 0.3)
llm_setting = req.get("llm_setting", {})
meta_data_filter = req.get("meta_data_filter", {})
prompt_config = req["prompt_config"]
if not is_create:
@ -85,6 +86,7 @@ def set_dialog():
"llm_id": llm_id,
"llm_setting": llm_setting,
"prompt_config": prompt_config,
"meta_data_filter": meta_data_filter,
"top_n": top_n,
"top_k": top_k,
"rerank_id": rerank_id,

View File

@ -681,6 +681,11 @@ def set_meta():
return get_json_result(data=False, message="No authorization.", code=settings.RetCode.AUTHENTICATION_ERROR)
try:
meta = json.loads(req["meta"])
if not isinstance(meta, dict):
return get_json_result(data=False, message="Only dictionary type supported.", code=settings.RetCode.ARGUMENT_ERROR)
for k,v in meta.items():
if not isinstance(v, str) and not isinstance(v, int) and not isinstance(v, float):
return get_json_result(data=False, message=f"The type is not supported: {v}", code=settings.RetCode.ARGUMENT_ERROR)
except Exception as e:
return get_json_result(data=False, message=f"Json syntax error: {e}", code=settings.RetCode.ARGUMENT_ERROR)
if not isinstance(meta, dict):

View File

@ -351,6 +351,7 @@ def knowledge_graph(kb_id):
obj["graph"]["edges"] = sorted(filtered_edges, key=lambda x: x.get("weight", 0), reverse=True)[:128]
return get_json_result(data=obj)
@manager.route('/<kb_id>/knowledge_graph', methods=['DELETE']) # noqa: F821
@login_required
def delete_knowledge_graph(kb_id):
@ -364,3 +365,17 @@ def delete_knowledge_graph(kb_id):
settings.docStoreConn.delete({"knowledge_graph_kwd": ["graph", "subgraph", "entity", "relation"]}, search.index_name(kb.tenant_id), kb_id)
return get_json_result(data=True)
@manager.route("/get_meta", methods=["GET"]) # noqa: F821
@login_required
def get_meta():
kb_ids = request.args.get("kb_ids", "").split(",")
for kb_id in kb_ids:
if not KnowledgebaseService.accessible(kb_id, current_user.id):
return get_json_result(
data=False,
message='No authorization.',
code=settings.RetCode.AUTHENTICATION_ERROR
)
return get_json_result(data=DocumentService.get_meta_by_kbs(kb_ids))