AI/ragflow
1
0
Fork 0
mirror of https://git.mirrors.martin98.com/https://github.com/infiniflow/ragflow.git synced 2025-04-30 15:55:53 +08:00
Code Issues Packages Projects Releases Wiki Activity

<think> tag is missing. (#7256)

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

Some models force thinking, resulting in the absence of the think tag in
the returned content

### Type of change

- [x] Bug Fix (non-breaking change which fixes an issue)
This commit is contained in:
WhiteBear 2025-04-24 11:44:10 +08:00 committed by GitHub
parent 4e8fd73a20
commit 2c62652ea8
No known key found for this signature in database
GPG Key ID: B5690EEEBB952194
8 changed files with 13 additions and 13 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
agent/component/exesql.py
View File

@ -61,7 +61,7 @@ class ExeSQL(Generate, ABC):
component_name = "ExeSQL"
def _refactor(self, ans):
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
match = re.search(r"```sql\s*(.*?)\s*```", ans, re.DOTALL)
if match:
ans = match.group(1) # Query content

2
agent/component/generate.py
View File

@ -200,7 +200,7 @@ class Generate(ComponentBase):
if len(msg) < 2:
msg.append({"role": "user", "content": "Output: "})
ans = chat_mdl.chat(msg[0]["content"], msg[1:], self._param.gen_conf())
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
if self._param.cite and "chunks" in retrieval_res.columns:
res = self.set_cite(retrieval_res, ans)

2
agent/component/keyword.py
View File

@ -57,7 +57,7 @@ class KeywordExtract(Generate, ABC):
ans = chat_mdl.chat(self._param.get_prompt(), [{"role": "user", "content": query}],
self._param.gen_conf())
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r".*keyword:", "", ans).strip()
logging.debug(f"ans: {ans}")
return KeywordExtract.be_output(ans)

4
agentic_reasoning/deep_research.py
View File

@ -57,7 +57,7 @@ class DeepResearcher:
msg_history[-1]["content"] += "\n\nContinues reasoning with the new information.\n"
for ans in self.chat_mdl.chat_streamly(REASON_PROMPT, msg_history, {"temperature": 0.7}):
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
if not ans:
continue
query_think = ans
@ -142,7 +142,7 @@ class DeepResearcher:
[{"role": "user",
"content": f'Now you should analyze each web page and find helpful information based on the current search query "{search_query}" and previous reasoning steps.'}],
{"temperature": 0.7}):
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
if not ans:
continue
summary_think = ans

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

@ -400,7 +400,7 @@ def chat(dialog, messages, stream=True, **kwargs):
answer = ""
for ans in chat_mdl.chat_streamly(prompt + prompt4citation, msg[1:], gen_conf):
if thought:
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
answer = ans
delta_ans = ans[len(last_ans) :]
if num_tokens_from_string(delta_ans) < 16:
@ -436,7 +436,7 @@ Please write the SQL, only SQL, without any other explanations or text.
def get_table():
nonlocal sys_prompt, user_prompt, question, tried_times
sql = chat_mdl.chat(sys_prompt, [{"role": "user", "content": user_prompt}], {"temperature": 0.06})
sql = re.sub(r"<think>.*</think>", "", sql, flags=re.DOTALL)
sql = re.sub(r"^.*</think>", "", sql, flags=re.DOTALL)
logging.debug(f"{question} ==> {user_prompt} get SQL: {sql}")
sql = re.sub(r"[\r\n]+", " ", sql.lower())
sql = re.sub(r".*select ", "select ", sql.lower())

2
graphrag/general/extractor.py
View File

@ -54,7 +54,7 @@ class Extractor:
return response
_, system_msg = message_fit_in([{"role": "system", "content": system}], int(self._llm.max_length * 0.92))
response = self._llm.chat(system_msg[0]["content"], hist, conf)
response = re.sub(r"<think>.*</think>", "", response, flags=re.DOTALL)
response = re.sub(r"^.*</think>", "", response, flags=re.DOTALL)
if response.find("**ERROR**") >= 0:
logging.warning(f"Extractor._chat got error. response: {response}")
return ""

8
rag/prompts.py
View File

@ -196,7 +196,7 @@ Requirements:
kwd = chat_mdl.chat(prompt, msg[1:], {"temperature": 0.2})
if isinstance(kwd, tuple):
kwd = kwd[0]
kwd = re.sub(r"<think>.*</think>", "", kwd, flags=re.DOTALL)
kwd = re.sub(r"^.*</think>", "", kwd, flags=re.DOTALL)
if kwd.find("**ERROR**") >= 0:
return ""
return kwd
@ -223,7 +223,7 @@ Requirements:
kwd = chat_mdl.chat(prompt, msg[1:], {"temperature": 0.2})
if isinstance(kwd, tuple):
kwd = kwd[0]
kwd = re.sub(r"<think>.*</think>", "", kwd, flags=re.DOTALL)
kwd = re.sub(r"^.*</think>", "", kwd, flags=re.DOTALL)
if kwd.find("**ERROR**") >= 0:
return ""
return kwd
@ -303,7 +303,7 @@ Output: What's the weather in Rochester on {tomorrow}?
###############
"""
ans = chat_mdl.chat(prompt, [{"role": "user", "content": "Output: "}], {"temperature": 0.2})
ans = re.sub(r"<think>.*</think>", "", ans, flags=re.DOTALL)
ans = re.sub(r"^.*</think>", "", ans, flags=re.DOTALL)
return ans if ans.find("**ERROR**") < 0 else messages[-1]["content"]
@ -350,7 +350,7 @@ Output:
kwd = chat_mdl.chat(prompt, msg[1:], {"temperature": 0.5})
if isinstance(kwd, tuple):
kwd = kwd[0]
kwd = re.sub(r"<think>.*</think>", "", kwd, flags=re.DOTALL)
kwd = re.sub(r"^.*</think>", "", kwd, flags=re.DOTALL)
if kwd.find("**ERROR**") >= 0:
raise Exception(kwd)

2
rag/raptor.py
View File

@ -48,7 +48,7 @@ class RecursiveAbstractiveProcessing4TreeOrganizedRetrieval:
response = await trio.to_thread.run_sync(
lambda: self._llm_model.chat(system, history, gen_conf)
)
response = re.sub(r"<think>.*</think>", "", response, flags=re.DOTALL)
response = re.sub(r"^.*</think>", "", response, flags=re.DOTALL)
if response.find("**ERROR**") >= 0:
raise Exception(response)
set_llm_cache(self._llm_model.llm_name, system, response, history, gen_conf)