BifrostX - Intelligent Assembling, Unlimited Possibilities!
Project description
BifrostX
插件式快速服务框架
安装
pip install bifrostx
快速开始
bifrostx server
开发指南
所有拓展必须在对应的目录下,且必须有__init__.py文件与bifrostx.toml文件
Interface
定义功能接口,解耦组件与适配器,应用与实现分离。
bifrostx.toml参数定义如下:
version="0.1.0" # 版本号
bifrostx_version="0.1.0" # 最低兼容bifrostx版本号,可选
display_name="大语言模型对话接口" # 接口描述名称
enter_class = "interface:Interface" # 入口类,必须继承 bifrostx.interface.BaseInterface, 可选默认为模块下 Interface类
Demo
from bifrostx.interface.base import BaseInterface
from abc import abstractmethod
# 定义接口
class Interface(BaseInterface):
@abstractmethod
def chat(
self,
prompt: List[ChatHistory],
temperature: Optional[confloat(gt=0, lt=1)] = None,
top_p: Optional[confloat(gt=0, lt=1)] = None,
) -> ChatHistory:
return NotImplemented
Adapter
适配器,实现Interface接口功能。
bifrostx.toml参数定义如下:
version = "0.1.0" # 版本号
bifrostx_version = "0.1.0" # 最低兼容bifrostx版本号,可选
display_name = "OpenAI-GPT" # 适配器描述名称
dependencies = ["openai", "tiktoken"] # 依赖的pypi包,可选
enter_class = "adapter:Adapter" # 入口类,必须继承要实现的 Interface接口,可选默认为模块下 Adapter类
[[implements]] # 实现的接口列表
interface = "llm_chat" # 实现接口名称
interface_version = "0.1.0" # 实现接口版本号
Demo
from bifrostx.config import Config
from bifrostx.utils.logger import logger
from bifrostx.core.data_model import BaseModel, validate_call, confloat
from Interfaces.llm_chat.interface import Interface
# 配置类
class AdapterInstanceConfig(BaseModel):
gpt_model: OpenaiModel = OpenaiModel.GPT35
api_base: str = "https://api.openai.com/v1"
api_type: OpenaiApiType = OpenaiApiType.OPENAI
api_key: str
default_temperature: confloat(gt=0, lt=1) = 0.2
# 适配器
class Adapter(Interface):
# 注册适配器实例配置
instance_config_schema = AdapterInstanceConfig
def __init__(self, instance_config: AdapterInstanceConfig):
super().__init__(instance_config)
# 从 config.toml 中获取配置
self.config = Config.get_extension_config(__name__)
# 参数验证装饰器
@validate_call
# 实现接口方法
def chat(
self,
prompt: List[ChatHistory],
temperature: Optional[confloat(gt=0, lt=1)] = None,
top_p: Optional[confloat(gt=0, lt=1)] = None,
):
inputs = ChatInputs(prompt=prompt, temperature=temperature, top_p=top_p)
resp = self._chat(inputs, use_stream=False)
return ChatHistory(**resp["choices"][0]["message"])
Component
组件,组合功能实现应用 api。服务接口默认都在/api路由下
bifrostx.toml参数定义如下:
display_name = "对话" # 组件描述名称
version = "0.1.0" # 版本号
bifrostx_version = "0.1.0" # 最低兼容bifrostx版本号,可选
enter_class = "component:Component" # 入口类,必须继承 bifrostx.component.BaseComponent, 可选默认为模块下 Component类
[[references]] # 引用的接口列表
interface = "llm_chat" # 引用接口名称
interface_version = "0.1.0" # 引用接口版本号
组件api定义必须以方法开头api_,api 请求参数与返回值,只能为json 可转换格式。
Demo
from bifrostx.component.base import BaseComponent
from Interfaces.llm_chat.interface import Interface as LLL_Chat_Interface
from bifrostx.core.data_model import BaseModel
# 组件实例配置类
class ComponentConfig(BaseModel):
llm_chat_instance: str
# api请求参数类
class ApiChatCompletionsInputs(BaseModel):
messages: List[ChatHistory]
# 组件
class Component(BaseComponent):
# 注册组件实例配置类
instance_config_schema = ComponentConfig
@property
def llm_chat_instance(self):
# 获取接口实现实例
instance = LLL_Chat_Interface.get_instance(
self.instance_config.llm_chat_instance
)
if instance is None:
raise Exception("LLM Chat instance not found")
return instance
# api接口chat_completions
def api_chat_completions(self, inputs: ApiChatCompletionsInputs):
# 调用接口方法
output = self.llm_chat_instance.chat(prompt=inputs.messages)
return output
Fontend
前端目录,静态资源目录默认为fontend, 默认访问首页为index.html
/api、/docs为保留路由, 同名目录不生效
配置
config.toml 实例配置
LOG_LEVEL: str = "DEBUG" # 日志级别
EXTENSION_DIR: str = "./extensions" # 扩展目录
FONTEND_DIR: str = "frontend" # 前端目录
[Adapters.llm_openai_gpt.config] # 适配器配置 如:适配器为 llm_openai_gpt
proxy = "http://xxxx.xxxx.xxxx"
[Adapters.llm_openai_gpt.instances.gpt] # 适配器实例配置 如:适配器为 llm_openai_gpt 实例名称 gpt
api_key = "sk-xxxxxxxxxxx"
server.toml 服务配置
app_name = "DemoServer" # 应用名称
server_bind = "0.0.0.0:18000" # 服务器绑定地址
[routers.chat_to_gpt] # 路由path 路由即为 /api/chat_to_gpt
component = "chat_with_llm" # 组件名称
summary = "聊天" # 路由描述
[routers.chat_to_gpt.config] # 组件配置
llm_chat_instance = "gpt"
[routers.chat_to_glm] # 路由配置
component = "chat_with_llm" # 组件名称
[routers.chat_to_glm.config] # 组件配置
llm_chat_instance = "glm"
Project details
Release history Release notifications | RSS feed
Download files
Download the file for your platform. If you're not sure which to choose, learn more about installing packages.
Source Distribution
Built Distribution
Filter files by name, interpreter, ABI, and platform.
If you're not sure about the file name format, learn more about wheel file names.
Copy a direct link to the current filters
File details
Details for the file bifrostx-0.1.21.tar.gz.
File metadata
- Download URL: bifrostx-0.1.21.tar.gz
- Upload date:
- Size: 11.4 kB
- Tags: Source
- Uploaded using Trusted Publishing? No
- Uploaded via: pdm/2.10.4 CPython/3.10.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
fc05c75fa71c99e95df6a3bc2a7f6fcfb8f1372787c527d990e7fd61c37bcda8
|
|
| MD5 |
d1a2091fffd4c329951f648d7a4db50f
|
|
| BLAKE2b-256 |
d938c1d3d2d6986cd18563125f400f74646757065c8cfc403504a910eb3a225a
|
File details
Details for the file bifrostx-0.1.21-py3-none-any.whl.
File metadata
- Download URL: bifrostx-0.1.21-py3-none-any.whl
- Upload date:
- Size: 15.1 kB
- Tags: Python 3
- Uploaded using Trusted Publishing? No
- Uploaded via: pdm/2.10.4 CPython/3.10.9
File hashes
| Algorithm | Hash digest | |
|---|---|---|
| SHA256 |
0b82a8a50cc00c25229fd05fbb5c3b5457b7619c3a30510d6c4fb323c59068b1
|
|
| MD5 |
b9986cbb5af525675dbd96a26eaae90b
|
|
| BLAKE2b-256 |
8ebe330ff6732164da17bb0b37af63730426b3b49127a26ce3272acb48046bf9
|