pengjunjie/ragflow
1
0
Fork 1
mirror of https://github.com/infiniflow/ragflow.git synced 2025-12-08 12:32:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

Feat: add or logic operations for meta data filters. (#11404)

Browse Source 
### What problem does this PR solve?

#11376 #11387

### Type of change

- [x] New Feature (non-breaking change which adds functionality)
This commit is contained in:
Kevin Hu
2025-11-20 14:31:12 +08:00
committed by GitHub
parent d2b1da0e26
commit 06cef71ba6
11 changed files with 129 additions and 48 deletions
Download Patch File Download Diff File Expand all files Collapse all files

6
api/apps/chunk_app.py
View File

@ -305,12 +305,12 @@ async def retrieval_test():
metas = DocumentService.get_meta_by_kbs(kb_ids)
if meta_data_filter.get("method") == "auto":
chat_mdl = LLMBundle(current_user.id, LLMType.CHAT, llm_name=search_config.get("chat_id", ""))
filters = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters))
filters: dict = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
if not doc_ids:
doc_ids = None
elif meta_data_filter.get("method") == "manual":
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
if not doc_ids:
doc_ids = None

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

@ -159,10 +159,10 @@ async def webhook(tenant_id: str, agent_id: str):
data=False, message=str(e),
code=RetCode.EXCEPTION_ERROR)
def sse():
async def sse():
nonlocal canvas
try:
for ans in canvas.run(query=req.get("query", ""), files=req.get("files", []), user_id=req.get("user_id", tenant_id), webhook_payload=req):
async for ans in canvas.run(query=req.get("query", ""), files=req.get("files", []), user_id=req.get("user_id", tenant_id), webhook_payload=req):
yield "data:" + json.dumps(ans, ensure_ascii=False) + "\n\n"
cvs.dsl = json.loads(str(canvas))

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

@ -120,7 +120,7 @@ async def retrieval(tenant_id):
retrieval_setting = req.get("retrieval_setting", {})
similarity_threshold = float(retrieval_setting.get("score_threshold", 0.0))
top = int(retrieval_setting.get("top_k", 1024))
metadata_condition = req.get("metadata_condition", {})
metadata_condition = req.get("metadata_condition", {}) or {}
metas = DocumentService.get_meta_by_kbs([kb_id])
doc_ids = []
@ -132,7 +132,7 @@ async def retrieval(tenant_id):
embd_mdl = LLMBundle(kb.tenant_id, LLMType.EMBEDDING.value, llm_name=kb.embd_id)
if metadata_condition:
doc_ids.extend(meta_filter(metas, convert_conditions(metadata_condition)))
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(

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

@ -1442,9 +1442,9 @@ async def retrieval_test(tenant_id):
if doc_id not in doc_ids_list:
return get_error_data_result(f"The datasets don't own the document {doc_id}")
if not doc_ids:
metadata_condition = req.get("metadata_condition", {})
metadata_condition = req.get("metadata_condition", {}) or {}
metas = DocumentService.get_meta_by_kbs(kb_ids)
doc_ids = meta_filter(metas, convert_conditions(metadata_condition))
doc_ids = meta_filter(metas, convert_conditions(metadata_condition), metadata_condition.get("logic", "and"))
similarity_threshold = float(req.get("similarity_threshold", 0.2))
vector_similarity_weight = float(req.get("vector_similarity_weight", 0.3))
top = int(req.get("top_k", 1024))

14
api/apps/sdk/session.py
View File

@ -428,17 +428,15 @@ async def agents_completion_openai_compatibility(tenant_id, agent_id):
return resp
else:
# For non-streaming, just return the response directly
response = next(
completion_openai(
async for response in completion_openai(
tenant_id,
agent_id,
question,
session_id=req.pop("session_id", req.get("id", "")) or req.get("metadata", {}).get("id", ""),
stream=False,
**req,
)
)
return jsonify(response)
):
return jsonify(response)
@manager.route("/agents/<agent_id>/completions", methods=["POST"]) # noqa: F821
@ -977,12 +975,12 @@ async def retrieval_test_embedded():
metas = DocumentService.get_meta_by_kbs(kb_ids)
if meta_data_filter.get("method") == "auto":
chat_mdl = LLMBundle(tenant_id, LLMType.CHAT, llm_name=search_config.get("chat_id", ""))
filters = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters))
filters: dict = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
if not doc_ids:
doc_ids = None
elif meta_data_filter.get("method") == "manual":
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
if not doc_ids:
doc_ids = None

10
api/db/services/canvas_service.py
View File

@ -177,7 +177,7 @@ class UserCanvasService(CommonService):
return True
def completion(tenant_id, agent_id, session_id=None, **kwargs):
async def completion(tenant_id, agent_id, session_id=None, **kwargs):
query = kwargs.get("query", "") or kwargs.get("question", "")
files = kwargs.get("files", [])
inputs = kwargs.get("inputs", {})
@ -219,7 +219,7 @@ def completion(tenant_id, agent_id, session_id=None, **kwargs):
"id": message_id
})
txt = ""
for ans in canvas.run(query=query, files=files, user_id=user_id, inputs=inputs):
async for ans in canvas.run(query=query, files=files, user_id=user_id, inputs=inputs):
ans["session_id"] = session_id
if ans["event"] == "message":
txt += ans["data"]["content"]
@ -237,7 +237,7 @@ def completion(tenant_id, agent_id, session_id=None, **kwargs):
API4ConversationService.append_message(conv["id"], conv)
def completion_openai(tenant_id, agent_id, question, session_id=None, stream=True, **kwargs):
async def completion_openai(tenant_id, agent_id, question, session_id=None, stream=True, **kwargs):
tiktoken_encoder = tiktoken.get_encoding("cl100k_base")
prompt_tokens = len(tiktoken_encoder.encode(str(question)))
user_id = kwargs.get("user_id", "")
@ -245,7 +245,7 @@ def completion_openai(tenant_id, agent_id, question, session_id=None, stream=Tru
if stream:
completion_tokens = 0
try:
for ans in completion(
async for ans in completion(
tenant_id=tenant_id,
agent_id=agent_id,
session_id=session_id,
@ -304,7 +304,7 @@ def completion_openai(tenant_id, agent_id, question, session_id=None, stream=Tru
try:
all_content = ""
reference = {}
for ans in completion(
async for ans in completion(
tenant_id=tenant_id,
agent_id=agent_id,
session_id=session_id,

25
api/db/services/dialog_service.py
View File

@ -287,7 +287,7 @@ def convert_conditions(metadata_condition):
]
def meta_filter(metas: dict, filters: list[dict]):
def meta_filter(metas: dict, filters: list[dict], logic: str = "and"):
doc_ids = set([])
def filter_out(v2docs, operator, value):
@ -331,7 +331,10 @@ def meta_filter(metas: dict, filters: list[dict]):
if not doc_ids:
doc_ids = set(ids)
else:
doc_ids = doc_ids & set(ids)
if logic == "and":
doc_ids = doc_ids & set(ids)
else:
doc_ids = doc_ids | set(ids)
if not doc_ids:
return []
return list(doc_ids)
@ -407,12 +410,12 @@ def chat(dialog, messages, stream=True, **kwargs):
if dialog.meta_data_filter:
metas = DocumentService.get_meta_by_kbs(dialog.kb_ids)
if dialog.meta_data_filter.get("method") == "auto":
filters = gen_meta_filter(chat_mdl, metas, questions[-1])
attachments.extend(meta_filter(metas, filters))
filters: dict = gen_meta_filter(chat_mdl, metas, questions[-1])
attachments.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
if not attachments:
attachments = None
elif dialog.meta_data_filter.get("method") == "manual":
attachments.extend(meta_filter(metas, dialog.meta_data_filter["manual"]))
attachments.extend(meta_filter(metas, dialog.meta_data_filter["manual"], dialog.meta_data_filter.get("logic", "and")))
if not attachments:
attachments = None
@ -778,12 +781,12 @@ def ask(question, kb_ids, tenant_id, chat_llm_name=None, search_config={}):
if meta_data_filter:
metas = DocumentService.get_meta_by_kbs(kb_ids)
if meta_data_filter.get("method") == "auto":
filters = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters))
filters: dict = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
if not doc_ids:
doc_ids = None
elif meta_data_filter.get("method") == "manual":
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
if not doc_ids:
doc_ids = None
@ -853,12 +856,12 @@ def gen_mindmap(question, kb_ids, tenant_id, search_config={}):
if meta_data_filter:
metas = DocumentService.get_meta_by_kbs(kb_ids)
if meta_data_filter.get("method") == "auto":
filters = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters))
filters: dict = gen_meta_filter(chat_mdl, metas, question)
doc_ids.extend(meta_filter(metas, filters["conditions"], filters.get("logic", "and")))
if not doc_ids:
doc_ids = None
elif meta_data_filter.get("method") == "manual":
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"]))
doc_ids.extend(meta_filter(metas, meta_data_filter["manual"], meta_data_filter.get("logic", "and")))
if not doc_ids:
doc_ids = None