pengjunjie/ragflow
1
0
Fork 1
mirror of https://github.com/infiniflow/ragflow.git synced 2025-12-26 17:16:52 +08:00
Code Issues Packages Projects Releases Wiki Activity

Fix: metadata issue & graphrag speeding up. (#12113)

Browse Source 
### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)

---------

Co-authored-by: Liu An <asiro@qq.com>
This commit is contained in:
Kevin Hu
2025-12-23 15:57:27 +08:00
committed by GitHub
parent 063b06494a
commit 00bb6fbd28
6 changed files with 6 additions and 11 deletions
Download Patch File Download Diff File Expand all files Collapse all files

4
api/db/services/dialog_service.py
View File

@ -406,7 +406,7 @@ async def async_chat(dialog, messages, stream=True, **kwargs):
dialog.vector_similarity_weight, dialog.vector_similarity_weight,
doc_ids=attachments, doc_ids=attachments,
top=dialog.top_k, top=dialog.top_k,
aggs=False, aggs=True,
rerank_mdl=rerank_mdl, rerank_mdl=rerank_mdl,
rank_feature=label_question(" ".join(questions), kbs), rank_feature=label_question(" ".join(questions), kbs),
) )
@ -769,7 +769,7 @@ async def async_ask(question, kb_ids, tenant_id, chat_llm_name=None, search_conf
vector_similarity_weight=search_config.get("vector_similarity_weight", 0.3), vector_similarity_weight=search_config.get("vector_similarity_weight", 0.3),
top=search_config.get("top_k", 1024), top=search_config.get("top_k", 1024),
doc_ids=doc_ids, doc_ids=doc_ids,
aggs=False, aggs=True,
rerank_mdl=rerank_mdl, rerank_mdl=rerank_mdl,
rank_feature=label_question(question, kbs) rank_feature=label_question(question, kbs)
) )

2
api/db/services/tenant_llm_service.py
View File

@ -97,7 +97,7 @@ class TenantLLMService(CommonService):
if llm_type == LLMType.EMBEDDING.value: if llm_type == LLMType.EMBEDDING.value:
mdlnm = tenant.embd_id if not llm_name else llm_name mdlnm = tenant.embd_id if not llm_name else llm_name
elif llm_type == LLMType.SPEECH2TEXT.value: elif llm_type == LLMType.SPEECH2TEXT.value:
mdlnm = tenant.asr_id mdlnm = tenant.asr_id if not llm_name else llm_name
elif llm_type == LLMType.IMAGE2TEXT.value: elif llm_type == LLMType.IMAGE2TEXT.value:
mdlnm = tenant.img2txt_id if not llm_name else llm_name mdlnm = tenant.img2txt_id if not llm_name else llm_name
elif llm_type == LLMType.CHAT.value: elif llm_type == LLMType.CHAT.value:

3
graphrag/general/extractor.py
View File

@ -71,18 +71,17 @@ class Extractor:
_, system_msg = message_fit_in([{"role": "system", "content": system}], int(self._llm.max_length * 0.92)) _, system_msg = message_fit_in([{"role": "system", "content": system}], int(self._llm.max_length * 0.92))
response = "" response = ""
for attempt in range(3): for attempt in range(3):
if task_id: if task_id:
if has_canceled(task_id): if has_canceled(task_id):
logging.info(f"Task {task_id} cancelled during entity resolution candidate processing.") logging.info(f"Task {task_id} cancelled during entity resolution candidate processing.")
raise TaskCanceledException(f"Task {task_id} was cancelled") raise TaskCanceledException(f"Task {task_id} was cancelled")
try: try:
response = asyncio.run(self._llm.async_chat(system_msg[0]["content"], hist, conf)) response = asyncio.run(self._llm.async_chat(system_msg[0]["content"], hist, conf))
response = re.sub(r"^.*</think>", "", response, flags=re.DOTALL) response = re.sub(r"^.*</think>", "", response, flags=re.DOTALL)
if response.find("**ERROR**") >= 0: if response.find("**ERROR**") >= 0:
raise Exception(response) raise Exception(response)
set_llm_cache(self._llm.llm_name, system, response, history, gen_conf) set_llm_cache(self._llm.llm_name, system, response, history, gen_conf)
break
except Exception as e: except Exception as e:
logging.exception(e) logging.exception(e)
if attempt == 2: if attempt == 2:

2
graphrag/general/index.py
View File

@ -198,7 +198,7 @@ async def run_graphrag_for_kb(
for d in raw_chunks: for d in raw_chunks:
content = d["content_with_weight"] content = d["content_with_weight"]
if num_tokens_from_string(current_chunk + content) < 1024: if num_tokens_from_string(current_chunk + content) < 4096:
current_chunk += content current_chunk += content
else: else:
if current_chunk: if current_chunk:

4
graphrag/light/graph_extractor.py
View File

@ -78,10 +78,6 @@ class GraphExtractor(Extractor):
hint_prompt = self._entity_extract_prompt.format(**self._context_base, input_text=content) hint_prompt = self._entity_extract_prompt.format(**self._context_base, input_text=content)
gen_conf = {} gen_conf = {}
final_result = ""
glean_result = ""
if_loop_result = ""
history = []
logging.info(f"Start processing for {chunk_key}: {content[:25]}...") logging.info(f"Start processing for {chunk_key}: {content[:25]}...")
if self.callback: if self.callback:
self.callback(msg=f"Start processing for {chunk_key}: {content[:25]}...") self.callback(msg=f"Start processing for {chunk_key}: {content[:25]}...")

2
rag/svr/task_executor.py
View File

@ -380,7 +380,7 @@ async def build_chunks(task, progress_callback):
cached = await gen_metadata(chat_mdl, cached = await gen_metadata(chat_mdl,
metadata_schema(task["parser_config"]["metadata"]), metadata_schema(task["parser_config"]["metadata"]),
d["content_with_weight"]) d["content_with_weight"])
set_llm_cache(chat_mdl.llm_name, d["content_with_weight"], cached, "metadata") set_llm_cache(chat_mdl.llm_name, d["content_with_weight"], cached, "metadata", {})
if cached: if cached:
d["metadata_obj"] = cached d["metadata_obj"] = cached
tasks = [] tasks = []