mirror of
https://github.com/RYDE-WORK/Langchain-Chatchat.git
synced 2026-01-19 21:37:20 +08:00
b4c68ddd05
* 优化在线 API ，支持 completion 和 embedding，简化在线 API 开发方式 新功能 - 智谱AI、Minimax、千帆、千问 4 个在线模型支持 embeddings（不通过Fastchat，后续会单独提供相关api接口） - 在线模型自动检测传入参数，在传入非 messages 格式的 prompt 时，自动转换为 completion 形式，以支持 completion 接口 开发者： - 重构ApiModelWorker: - 所有在线 API 请求封装到 do_chat 方法：自动传入参数 ApiChatParams，简化参数与配置项的获取；自动处理与fastchat的接口 - 加强 API 请求错误处理，返回更有意义的信息 - 改用 qianfan sdk 重写 qianfan-api - 将所有在线模型的测试用例统一在一起，简化测试用例编写 * Delete requirements_langflow.txt
102 lines
3.3 KiB
Python
102 lines
3.3 KiB
Python
import json
|
|
import sys
|
|
|
|
from fastchat.conversation import Conversation
|
|
from configs import TEMPERATURE
|
|
from http import HTTPStatus
|
|
from typing import List, Literal, Dict
|
|
|
|
from fastchat import conversation as conv
|
|
from server.model_workers.base import *
|
|
from server.model_workers.base import ApiEmbeddingsParams
|
|
|
|
|
|
class QwenWorker(ApiModelWorker):
|
|
DEFAULT_EMBED_MODEL = "text-embedding-v1"
|
|
|
|
def __init__(
|
|
self,
|
|
*,
|
|
version: Literal["qwen-turbo", "qwen-plus"] = "qwen-turbo",
|
|
model_names: List[str] = ["qwen-api"],
|
|
controller_addr: str = None,
|
|
worker_addr: str = None,
|
|
**kwargs,
|
|
):
|
|
kwargs.update(model_names=model_names, controller_addr=controller_addr, worker_addr=worker_addr)
|
|
kwargs.setdefault("context_len", 16384)
|
|
super().__init__(**kwargs)
|
|
self.version = version
|
|
|
|
def do_chat(self, params: ApiChatParams) -> Dict:
|
|
import dashscope
|
|
params.load_config(self.model_names[0])
|
|
|
|
gen = dashscope.Generation()
|
|
responses = gen.call(
|
|
model=params.version,
|
|
temperature=params.temperature,
|
|
api_key=params.api_key,
|
|
messages=params.messages,
|
|
result_format='message', # set the result is message format.
|
|
stream=True,
|
|
)
|
|
|
|
for resp in responses:
|
|
if resp["status_code"] == 200:
|
|
if choices := resp["output"]["choices"]:
|
|
yield {
|
|
"error_code": 0,
|
|
"text": choices[0]["message"]["content"],
|
|
}
|
|
else:
|
|
yield {
|
|
"error_code": resp["status_code"],
|
|
"text": resp["message"],
|
|
}
|
|
|
|
def do_embeddings(self, params: ApiEmbeddingsParams) -> Dict:
|
|
import dashscope
|
|
params.load_config(self.model_names[0])
|
|
|
|
resp = dashscope.TextEmbedding.call(
|
|
model=params.embed_model or self.DEFAULT_EMBED_MODEL,
|
|
input=params.texts, # 最大25行
|
|
api_key=params.api_key,
|
|
)
|
|
if resp["status_code"] != 200:
|
|
return {"code": resp["status_code"], "msg": resp.message}
|
|
else:
|
|
embeddings = [x["embedding"] for x in resp["output"]["embeddings"]]
|
|
return {"code": 200, "embeddings": embeddings}
|
|
|
|
def get_embeddings(self, params):
|
|
# TODO: 支持embeddings
|
|
print("embedding")
|
|
print(params)
|
|
|
|
def make_conv_template(self, conv_template: str = None, model_path: str = None) -> Conversation:
|
|
# TODO: 确认模板是否需要修改
|
|
return conv.Conversation(
|
|
name=self.model_names[0],
|
|
system_message="你是一个聪明、对人类有帮助的人工智能，你可以对人类提出的问题给出有用、详细、礼貌的回答。",
|
|
messages=[],
|
|
roles=["user", "assistant", "system"],
|
|
sep="\n### ",
|
|
stop_str="###",
|
|
)
|
|
|
|
|
|
if __name__ == "__main__":
|
|
import uvicorn
|
|
from server.utils import MakeFastAPIOffline
|
|
from fastchat.serve.model_worker import app
|
|
|
|
worker = QwenWorker(
|
|
controller_addr="http://127.0.0.1:20001",
|
|
worker_addr="http://127.0.0.1:20007",
|
|
)
|
|
sys.modules["fastchat.serve.model_worker"].worker = worker
|
|
MakeFastAPIOffline(app)
|
|
uvicorn.run(app, port=20007)
|