pengjunjie/ragflow
1
0
Fork 1
mirror of https://github.com/infiniflow/ragflow.git synced 2025-12-08 20:42:30 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix: cite disfunction for G component. (#7117)

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

#7097

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
This commit is contained in:
Kevin Hu
2025-04-18 18:05:26 +08:00
committed by GitHub
parent 8b8a2f2949
commit 487aed419e
3 changed files with 19 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

22
agent/component/generate.py
View File

@ -13,6 +13,7 @@
# See the License for the specific language governing permissions and
# limitations under the License.
#
import json
import re
from functools import partial
import pandas as pd
@ -74,29 +75,30 @@ class Generate(ComponentBase):
return list(cpnts)
def set_cite(self, retrieval_res, answer):
retrieval_res = retrieval_res.dropna(subset=["vector", "content_ltks"]).reset_index(drop=True)
if "empty_response" in retrieval_res.columns:
retrieval_res["empty_response"].fillna("", inplace=True)
chunks = json.loads(retrieval_res["chunks"][0])
answer, idx = settings.retrievaler.insert_citations(answer,
[ck["content_ltks"] for _, ck in retrieval_res.iterrows()],
[ck["vector"] for _, ck in retrieval_res.iterrows()],
[ck["content_ltks"] for ck in chunks],
[ck["vector"] for ck in chunks],
LLMBundle(self._canvas.get_tenant_id(), LLMType.EMBEDDING,
self._canvas.get_embedding_model()), tkweight=0.7,
vtweight=0.3)
doc_ids = set([])
recall_docs = []
for i in idx:
did = retrieval_res.loc[int(i), "doc_id"]
did = chunks[int(i)]["doc_id"]
if did in doc_ids:
continue
doc_ids.add(did)
recall_docs.append({"doc_id": did, "doc_name": retrieval_res.loc[int(i), "docnm_kwd"]})
recall_docs.append({"doc_id": did, "doc_name": chunks[int(i)]["docnm_kwd"]})
del retrieval_res["vector"]
del retrieval_res["content_ltks"]
for c in chunks:
del c["vector"]
del c["content_ltks"]
reference = {
"chunks": [ck.to_dict() for _, ck in retrieval_res.iterrows()],
"chunks": chunks,
"doc_aggs": recall_docs
}
@ -200,7 +202,7 @@ class Generate(ComponentBase):
ans = chat_mdl.chat(msg[0]["content"], msg[1:], self._param.gen_conf())
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
if self._param.cite and "content_ltks" in retrieval_res.columns and "vector" in retrieval_res.columns:
if self._param.cite and "chunks" in retrieval_res.columns:
res = self.set_cite(retrieval_res, ans)
return pd.DataFrame([res])
@ -229,7 +231,7 @@ class Generate(ComponentBase):
answer = ans
yield res
if self._param.cite and "content_ltks" in retrieval_res.columns and "vector" in retrieval_res.columns:
if self._param.cite and "chunks" in retrieval_res.columns:
res = self.set_cite(retrieval_res, answer)
yield res