aidlab/AiDA_Python
6
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

重构LLM构建逻辑，调整环境判断以支持本地和生产模式

Browse Source
This commit is contained in:
zcr
2026-06-17 12:14:14 +08:00
parent 456298cee7
commit f22d8d6013
1 changed files with 19 additions and 26 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
app/service/fashion_agent/init_llm.py
View File

@@ -11,7 +11,15 @@ SERVE_ENV = os.environ.get("SERVE_ENV", "dev") # prod 使用 Qwen API，dev 使
def build_llm(enable_thinking: bool = False): def build_llm(enable_thinking: bool = False):
if SERVE_ENV == "prod": if SERVE_ENV == "local":
llm = ChatOllama(
model="qwen3.6:27b",
base_url="http://127.0.0.1:11434",
reasoning=enable_thinking,
temperature=0.7,
num_ctx=32768,
)
else:
llm = ChatQwen( llm = ChatQwen(
model="qwen3.6-plus", model="qwen3.6-plus",
timeout=None, timeout=None,
@@ -20,31 +28,20 @@ def build_llm(enable_thinking: bool = False):
streaming=True, streaming=True,
api_key=QWEN_API_KEY_INTL, api_key=QWEN_API_KEY_INTL,
) )
else:
llm = ChatOllama(
model="qwen3.6:27b",
base_url="http://127.0.0.1:11434",
reasoning=enable_thinking,
temperature=0.7,
num_ctx=32768,
)
return llm return llm
# qwen_plus_llm = ChatQwen( if SERVE_ENV == "local":
# model="qwen-plus", qwen_plus_llm = ChatOllama(
# timeout=None, model="qwen3.6:27b",
# max_retries=2, base_url="http://127.0.0.1:11434",
# streaming=False, reasoning=False,
# temperature=0.25, temperature=0.7,
# top_p=0.8, num_ctx=32768,
# api_key=QWEN_API_KEY_INTL, extra_kwargs={"stream": False},
# ) )
# response = qwen_plus_llm.invoke("你好")
# print(response)
else:
if SERVE_ENV == "prod":
qwen_plus_llm = ChatQwen( qwen_plus_llm = ChatQwen(
model="qwen-plus", model="qwen-plus",
timeout=None, timeout=None,
@@ -54,9 +51,5 @@ if SERVE_ENV == "prod":
top_p=0.8, top_p=0.8,
api_key=QWEN_API_KEY_INTL, api_key=QWEN_API_KEY_INTL,
) )
else:
qwen_plus_llm = ChatOllama(
model="qwen3.6:27b", base_url="http://127.0.0.1:11434", reasoning=False, temperature=0.7, num_ctx=32768, extra_kwargs={"stream": False}
)
# for chunk in qwen_plus_llm.stream("你好"): # for chunk in qwen_plus_llm.stream("你好"):
# print(chunk) # print(chunk)