AI_POC/TERES_fastapi_backend
12
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

v0.21.1-fastapi

Browse Source
This commit is contained in:
dangzerong
2025-11-04 16:06:36 +08:00
parent 3e58c3d0e9
commit d57b5d76ae
218 changed files with 19617 additions and 72339 deletions
Download Patch File Download Diff File Expand all files Collapse all files

31
api/db/services/llm_service.py
View File

@@ -205,32 +205,31 @@ class LLMBundle(LLM4Tenant):
return txt
return txt[last_think_end + len("</think>") :]
@staticmethod
def _clean_param(chat_partial, **kwargs):
func = chat_partial.func
sig = inspect.signature(func)
keyword_args = []
support_var_args = False
for param in sig.parameters.values():
if param.kind == inspect.Parameter.VAR_KEYWORD or param.kind == inspect.Parameter.VAR_POSITIONAL:
support_var_args = True
elif param.kind == inspect.Parameter.KEYWORD_ONLY:
keyword_args.append(param.name)
allowed_params = set()
use_kwargs = kwargs
if not support_var_args:
use_kwargs = {k: v for k, v in kwargs.items() if k in keyword_args}
return use_kwargs
for param in sig.parameters.values():
if param.kind == inspect.Parameter.VAR_KEYWORD:
support_var_args = True
elif param.kind in (inspect.Parameter.POSITIONAL_OR_KEYWORD, inspect.Parameter.KEYWORD_ONLY):
allowed_params.add(param.name)
if support_var_args:
return kwargs
else:
return {k: v for k, v in kwargs.items() if k in allowed_params}
def chat(self, system: str, history: list, gen_conf: dict = {}, **kwargs) -> str:
if self.langfuse:
generation = self.langfuse.start_generation(trace_context=self.trace_context, name="chat", model=self.llm_name, input={"system": system, "history": history})
chat_partial = partial(self.mdl.chat, system, history, gen_conf)
chat_partial = partial(self.mdl.chat, system, history, gen_conf, **kwargs)
if self.is_tools and self.mdl.is_tools:
chat_partial = partial(self.mdl.chat_with_tools, system, history, gen_conf)
chat_partial = partial(self.mdl.chat_with_tools, system, history, gen_conf, **kwargs)
use_kwargs = self._clean_param(chat_partial, **kwargs)
txt, used_tokens = chat_partial(**use_kwargs)
txt = self._remove_reasoning_content(txt)
@@ -266,7 +265,7 @@ class LLMBundle(LLM4Tenant):
break
if txt.endswith("</think>"):
ans = ans.rstrip("</think>")
ans = ans[: -len("</think>")]
if not self.verbose_tool_use:
txt = re.sub(r"<tool_call>.*?</tool_call>", "", txt, flags=re.DOTALL)