Skip to main content

BifrostX - Intelligent Assembling, Unlimited Possibilities!

Project description

Python Version

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


Download files

Download the file for your platform. If you're not sure which to choose, learn more about installing packages.

Source Distribution

bifrostx-0.1.21.tar.gz (11.4 kB view details)

Uploaded Source

Built Distribution

If you're not sure about the file name format, learn more about wheel file names.

bifrostx-0.1.21-py3-none-any.whl (15.1 kB view details)

Uploaded Python 3

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

Hashes for bifrostx-0.1.21.tar.gz
Algorithm Hash digest
SHA256 fc05c75fa71c99e95df6a3bc2a7f6fcfb8f1372787c527d990e7fd61c37bcda8
MD5 d1a2091fffd4c329951f648d7a4db50f
BLAKE2b-256 d938c1d3d2d6986cd18563125f400f74646757065c8cfc403504a910eb3a225a

See more details on using hashes here.

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

Hashes for bifrostx-0.1.21-py3-none-any.whl
Algorithm Hash digest
SHA256 0b82a8a50cc00c25229fd05fbb5c3b5457b7619c3a30510d6c4fb323c59068b1
MD5 b9986cbb5af525675dbd96a26eaae90b
BLAKE2b-256 8ebe330ff6732164da17bb0b37af63730426b3b49127a26ce3272acb48046bf9

See more details on using hashes here.

Supported by

AWS Cloud computing and Security Sponsor Datadog Monitoring Depot Continuous Integration Fastly CDN Google Download Analytics Pingdom Monitoring Sentry Error logging StatusPage Status page