15308555518/ MaxKB
1
0
Fork 0
mirror of https://github.com/1Panel-dev/MaxKB.git synced 2025-12-26 01:33:05 +00:00
Code Issues Actions 2 Packages Projects Releases Wiki Activity

Merge upstream changes

This commit is contained in:
tongque 2024-04-28 13:36:09 +08:00
parent 601af09d03 d1fe18166a
commit 6f15ebba05
17 changed files with 263 additions and 100 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

24
apps/common/config/tokenizer_manage_config.py Normal file
View File

@ -0,0 +1,24 @@
# coding=utf-8
"""
@project: maxkb
@Author：虎
@file： tokenizer_manage_config.py
@date：2024/4/28 10:17
@desc:
"""
class TokenizerManage:
tokenizer = None
@staticmethod
def get_tokenizer():
from transformers import GPT2TokenizerFast
if TokenizerManage.tokenizer is None:
TokenizerManage.tokenizer = GPT2TokenizerFast.from_pretrained(
'gpt2',
cache_dir="/opt/maxkb/model/tokenizer",
local_files_only=True,
resume_download=False,
force_download=False)
return TokenizerManage.tokenizer

5
apps/setting/models_provider/impl/azure_model_provider/azure_model_provider.py
View File

@ -19,6 +19,7 @@ from common.util.file_util import get_file_content
from setting.models_provider.base_model_provider import IModelProvider, ModelProvideInfo, BaseModelCredential, \
ModelInfo, \
ModelTypeConst, ValidCode
from setting.models_provider.impl.azure_model_provider.model.azure_chat_model import AzureChatModel
from smartdoc.conf import PROJECT_DIR
@ -119,8 +120,8 @@ class AzureModelProvider(IModelProvider):
def get_model(self, model_type, model_name, model_credential: Dict[str, object], **model_kwargs) -> AzureChatOpenAI:
model_info: ModelInfo = model_dict.get(model_name)
azure_chat_open_ai = AzureChatOpenAI(
openai_api_base=model_credential.get('api_base'),
azure_chat_open_ai = AzureChatModel(
azure_endpoint=model_credential.get('api_base'),
openai_api_version=model_info.api_version if model_name in model_dict else model_credential.get(
'api_version'),
deployment_name=model_credential.get('deployment_name'),

24
apps/setting/models_provider/impl/azure_model_provider/model/azure_chat_model.py Normal file
View File

@ -0,0 +1,24 @@
# coding=utf-8
"""
@project: maxkb
@Author：虎
@file： azure_chat_model.py
@date：2024/4/28 11:45
@desc:
"""
from typing import List
from langchain_core.messages import BaseMessage, get_buffer_string
from langchain_openai import AzureChatOpenAI
from common.config.tokenizer_manage_config import TokenizerManage
class AzureChatModel(AzureChatOpenAI):
def get_num_tokens_from_messages(self, messages: List[BaseMessage]) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return sum([len(tokenizer.encode(get_buffer_string([m]))) for m in messages])
def get_num_tokens(self, text: str) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return len(tokenizer.encode(text))

14
apps/setting/models_provider/impl/kimi_model_provider/model/kimi_chat_model.py
View File

@ -11,19 +11,7 @@ from typing import List
from langchain_community.chat_models import ChatOpenAI
from langchain_core.messages import BaseMessage, get_buffer_string
class TokenizerManage:
tokenizer = None
@staticmethod
def get_tokenizer():
from transformers import GPT2TokenizerFast
if TokenizerManage.tokenizer is None:
TokenizerManage.tokenizer = GPT2TokenizerFast.from_pretrained('gpt2',
cache_dir="/opt/maxkb/model/tokenizer",
resume_download=False,
force_download=False)
return TokenizerManage.tokenizer
from common.config.tokenizer_manage_config import TokenizerManage
class KimiChatModel(ChatOpenAI):

14
apps/setting/models_provider/impl/ollama_model_provider/model/ollama_chat_model.py
View File

@ -11,19 +11,7 @@ from typing import List
from langchain_community.chat_models import ChatOpenAI
from langchain_core.messages import BaseMessage, get_buffer_string
class TokenizerManage:
tokenizer = None
@staticmethod
def get_tokenizer():
from transformers import GPT2TokenizerFast
if TokenizerManage.tokenizer is None:
TokenizerManage.tokenizer = GPT2TokenizerFast.from_pretrained('gpt2',
cache_dir="/opt/maxkb/model/tokenizer",
resume_download=False,
force_download=False)
return TokenizerManage.tokenizer
from common.config.tokenizer_manage_config import TokenizerManage
class OllamaChatModel(ChatOpenAI):

44
apps/setting/models_provider/impl/ollama_model_provider/ollama_model_provider.py
View File

@ -74,9 +74,49 @@ model_dict = {
'llama2-chinese:13b',
'由于Llama2本身的中文对齐较弱，我们采用中文指令集，对meta-llama/Llama-2-13b-chat-hf进行LoRA微调，使其具备较强的中文对话能力。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen-14b': ModelInfo(
'llama3:8b': ModelInfo(
'llama3:8b',
'Meta Llama 3：迄今为止最有能力的公开产品LLM。8亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'llama3:70b': ModelInfo(
'llama3:70b',
'Meta Llama 3：迄今为止最有能力的公开产品LLM。70亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:0.5b': ModelInfo(
'qwen:0.5b',
'qwen 1.5 0.5b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。0.5亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:1.8b': ModelInfo(
'qwen:1.8b',
'qwen 1.5 1.8b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。1.8亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:4b': ModelInfo(
'qwen:4b',
'qwen 1.5 4b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。4亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:7b': ModelInfo(
'qwen:7b',
'qwen 1.5 7b 相较于以往版本，模型与人类偏好的对齐程度以及多语1言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。7亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:14b': ModelInfo(
'qwen:14b',
'qwen 1.5 14b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。',
'qwen 1.5 14b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。14亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:32b': ModelInfo(
'qwen:32b',
'qwen 1.5 32b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。32亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:72b': ModelInfo(
'qwen:72b',
'qwen 1.5 72b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。72亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'qwen:110b': ModelInfo(
'qwen:110b',
'qwen 1.5 110b 相较于以往版本，模型与人类偏好的对齐程度以及多语言处理能力上有显著增强。所有规模的模型都支持32768个tokens的上下文长度。110亿参数。',
ModelTypeConst.LLM, ollama_llm_model_credential),
'phi3': ModelInfo(
'phi3',
'Phi-3 Mini是Microsoft的3.8B参数，轻量级，最先进的开放模型。',
ModelTypeConst.LLM, ollama_llm_model_credential),
}

14
apps/setting/models_provider/impl/openai_model_provider/model/openai_chat_model.py
View File

@ -11,19 +11,7 @@ from typing import List
from langchain_core.messages import BaseMessage, get_buffer_string
from langchain_openai import ChatOpenAI
class TokenizerManage:
tokenizer = None
@staticmethod
def get_tokenizer():
from transformers import GPT2TokenizerFast
if TokenizerManage.tokenizer is None:
TokenizerManage.tokenizer = GPT2TokenizerFast.from_pretrained('gpt2',
cache_dir="/opt/maxkb/model/tokenizer",
resume_download=False,
force_download=False)
return TokenizerManage.tokenizer
from common.config.tokenizer_manage_config import TokenizerManage
class OpenAIChatModel(ChatOpenAI):

35
apps/setting/models_provider/impl/openai_model_provider/openai_model_provider.py
View File

@ -57,12 +57,39 @@ class OpenAILLMModelCredential(BaseForm, BaseModelCredential):
openai_llm_model_credential = OpenAILLMModelCredential()
model_dict = {
'gpt-3.5-turbo': ModelInfo('gpt-3.5-turbo', '', ModelTypeConst.LLM, openai_llm_model_credential,
'gpt-3.5-turbo': ModelInfo('gpt-3.5-turbo', '最新的gpt-3.5-turbo，随OpenAI调整而更新', ModelTypeConst.LLM,
openai_llm_model_credential,
),
'gpt-3.5-turbo-0613': ModelInfo('gpt-3.5-turbo-0613', '', ModelTypeConst.LLM, openai_llm_model_credential,
'gpt-3.5-turbo-0125': ModelInfo('gpt-3.5-turbo-0125',
'2024年1月25日的gpt-3.5-turbo快照，支持上下文长度16,385 tokens', ModelTypeConst.LLM,
openai_llm_model_credential,
),
'gpt-3.5-turbo-1106': ModelInfo('gpt-3.5-turbo-1106',
'2023年11月6日的gpt-3.5-turbo快照，支持上下文长度16,385 tokens', ModelTypeConst.LLM,
openai_llm_model_credential,
),
'gpt-3.5-turbo-0613': ModelInfo('gpt-3.5-turbo-0613',
'[Legacy] 2023年6月13日的gpt-3.5-turbo快照，将于2024年6月13日弃用',
ModelTypeConst.LLM, openai_llm_model_credential,
),
'gpt-4': ModelInfo('gpt-4', '最新的gpt-4，随OpenAI调整而更新', ModelTypeConst.LLM, openai_llm_model_credential,
),
'gpt-4-turbo': ModelInfo('gpt-4-turbo', '最新的gpt-4-turbo，随OpenAI调整而更新', ModelTypeConst.LLM,
openai_llm_model_credential,
),
'gpt-4-turbo-preview': ModelInfo('gpt-4-turbo-preview', '最新的gpt-4-turbo-preview，随OpenAI调整而更新',
ModelTypeConst.LLM, openai_llm_model_credential,
),
'gpt-4-turbo-2024-04-09': ModelInfo('gpt-4-turbo-2024-04-09',
'2024年4月9日的gpt-4-turbo快照，支持上下文长度128,000 tokens',
ModelTypeConst.LLM, openai_llm_model_credential,
),
'gpt-4-0125-preview': ModelInfo('gpt-4-0125-preview', '2024年1月25日的gpt-4-turbo快照，支持上下文长度128,000 tokens',
ModelTypeConst.LLM, openai_llm_model_credential,
),
'gpt-4-1106-preview': ModelInfo('gpt-4-1106-preview', '2023年11月6日的gpt-4-turbo快照，支持上下文长度128,000 tokens',
ModelTypeConst.LLM, openai_llm_model_credential,
),
'gpt-4': ModelInfo('gpt-4', '', ModelTypeConst.LLM, openai_llm_model_credential,
)
}

24
apps/setting/models_provider/impl/qwen_model_provider/model/qwen_chat_model.py Normal file
View File

@ -0,0 +1,24 @@
# coding=utf-8
"""
@project: maxkb
@Author：虎
@file： qwen_chat_model.py
@date：2024/4/28 11:44
@desc:
"""
from typing import List
from langchain_community.chat_models import ChatTongyi
from langchain_core.messages import BaseMessage, get_buffer_string
from common.config.tokenizer_manage_config import TokenizerManage
class QwenChatModel(ChatTongyi):
def get_num_tokens_from_messages(self, messages: List[BaseMessage]) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return sum([len(tokenizer.encode(get_buffer_string([m]))) for m in messages])
def get_num_tokens(self, text: str) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return len(tokenizer.encode(text))

3
apps/setting/models_provider/impl/qwen_model_provider/qwen_model_provider.py
View File

@ -18,6 +18,7 @@ from common.forms import BaseForm
from common.util.file_util import get_file_content
from setting.models_provider.base_model_provider import ModelProvideInfo, ModelTypeConst, BaseModelCredential, \
ModelInfo, IModelProvider, ValidCode
from setting.models_provider.impl.qwen_model_provider.model.qwen_chat_model import QwenChatModel
from smartdoc.conf import PROJECT_DIR
@ -66,7 +67,7 @@ class QwenModelProvider(IModelProvider):
return 3
def get_model(self, model_type, model_name, model_credential: Dict[str, object], **model_kwargs) -> ChatTongyi:
chat_tong_yi = ChatTongyi(
chat_tong_yi = QwenChatModel(
model_name=model_name,
dashscope_api_key=model_credential.get('api_key')
)

14
apps/setting/models_provider/impl/wenxin_model_provider/model/qian_fan_chat_model.py
View File

@ -18,19 +18,7 @@ from langchain.schema.output import ChatGenerationChunk
from langchain.schema.runnable import RunnableConfig
from langchain_community.chat_models import QianfanChatEndpoint
class TokenizerManage:
tokenizer = None
@staticmethod
def get_tokenizer():
from transformers import GPT2TokenizerFast
if TokenizerManage.tokenizer is None:
TokenizerManage.tokenizer = GPT2TokenizerFast.from_pretrained('gpt2',
cache_dir="/opt/maxkb/model/tokenizer",
resume_download=False,
force_download=False)
return TokenizerManage.tokenizer
from common.config.tokenizer_manage_config import TokenizerManage
class QianfanChatModel(QianfanChatEndpoint):

12
apps/setting/models_provider/impl/xf_model_provider/model/xf_chat_model.py
View File

@ -12,11 +12,21 @@ from typing import List, Optional, Any, Iterator
from langchain_community.chat_models import ChatSparkLLM
from langchain_community.chat_models.sparkllm import _convert_message_to_dict, _convert_delta_to_message_chunk
from langchain_core.callbacks import CallbackManagerForLLMRun
from langchain_core.messages import BaseMessage, AIMessageChunk
from langchain_core.messages import BaseMessage, AIMessageChunk, get_buffer_string
from langchain_core.outputs import ChatGenerationChunk
from common.config.tokenizer_manage_config import TokenizerManage
class XFChatSparkLLM(ChatSparkLLM):
def get_num_tokens_from_messages(self, messages: List[BaseMessage]) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return sum([len(tokenizer.encode(get_buffer_string([m]))) for m in messages])
def get_num_tokens(self, text: str) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return len(tokenizer.encode(text))
def _stream(
self,
messages: List[BaseMessage],

24
apps/setting/models_provider/impl/zhipu_model_provider/model/zhipu_chat_model.py Normal file
View File

@ -0,0 +1,24 @@
# coding=utf-8
"""
@project: maxkb
@Author：虎
@file： zhipu_chat_model.py
@date：2024/4/28 11:42
@desc:
"""
from typing import List
from langchain_community.chat_models import ChatZhipuAI
from langchain_core.messages import BaseMessage, get_buffer_string
from common.config.tokenizer_manage_config import TokenizerManage
class ZhipuChatModel(ChatZhipuAI):
def get_num_tokens_from_messages(self, messages: List[BaseMessage]) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return sum([len(tokenizer.encode(get_buffer_string([m]))) for m in messages])
def get_num_tokens(self, text: str) -> int:
tokenizer = TokenizerManage.get_tokenizer()
return len(tokenizer.encode(text))

3
apps/setting/models_provider/impl/zhipu_model_provider/zhipu_model_provider.py
View File

@ -18,6 +18,7 @@ from common.forms import BaseForm
from common.util.file_util import get_file_content
from setting.models_provider.base_model_provider import ModelProvideInfo, ModelTypeConst, BaseModelCredential, \
ModelInfo, IModelProvider, ValidCode
from setting.models_provider.impl.zhipu_model_provider.model.zhipu_chat_model import ZhipuChatModel
from smartdoc.conf import PROJECT_DIR
@ -66,7 +67,7 @@ class ZhiPuModelProvider(IModelProvider):
return 3
def get_model(self, model_type, model_name, model_credential: Dict[str, object], **model_kwargs) -> ChatZhipuAI:
zhipuai_chat = ChatZhipuAI(
zhipuai_chat = ZhipuChatModel(
temperature=0.5,
api_key=model_credential.get('api_key'),
model=model_name

52
ui/src/layout/components/top-bar/index.vue
View File

@ -12,25 +12,47 @@
</div>
<div class="flex-center avatar">
<el-tooltip effect="dark" :content="$t('layout.topbar.github')" placement="top">
<AppIcon iconName="app-github" class="cursor color-secondary mr-8 ml-8" style="font-size: 20px"
@click="toUrl('https://github.com/1Panel-dev/MaxKB')"></AppIcon>
<AppIcon
iconName="app-github"
class="cursor color-secondary mr-8 ml-8"
style="font-size: 20px"
@click="toUrl('https://github.com/1Panel-dev/MaxKB')"
></AppIcon>
</el-tooltip>
<el-tooltip effect="dark" :content="$t('layout.topbar.handbook')" placement="top">
<AppIcon iconName="app-reading" class="cursor color-secondary mr-8 ml-8" style="font-size: 20px"
@click="toUrl('https://github.com/1Panel-dev/MaxKB/wiki')"></AppIcon>
<AppIcon
iconName="app-reading"
class="cursor color-secondary mr-8 ml-8"
style="font-size: 20px"
@click="toUrl('https://github.com/1Panel-dev/MaxKB/wiki')"
></AppIcon>
</el-tooltip>
<el-tooltip effect="dark" :content="$t('layout.topbar.forum')" placement="top">
<AppIcon iconName="app-help" class="cursor color-secondary mr-8 ml-8" style="font-size: 20px"
@click="toUrl('https://bbs.fit2cloud.com/c/mk/11')"></AppIcon>
<AppIcon
iconName="app-help"
class="cursor color-secondary mr-8 ml-8"
style="font-size: 20px"
@click="toUrl('https://bbs.fit2cloud.com/c/mk/11')"
></AppIcon>
</el-tooltip>
<el-dropdown trigger="click" type="primary">
<el-dropdown v-if="false" trigger="click" type="primary">
<template #dropdown>
<el-dropdown-menu>
<el-dropdown-item v-for="(lang, index) in langList" :key="index" :value="lang.value"
@click="changeLang(lang.value)">{{ lang.label }}</el-dropdown-item>
<el-dropdown-item
v-for="(lang, index) in langList"
:key="index"
:value="lang.value"
@click="changeLang(lang.value)"
>{{ lang.label }}</el-dropdown-item
>
</el-dropdown-menu>
</template>
<AppIcon iconName="app-translate" class="cursor color-secondary mr-16 ml-8" style="font-size: 20px" @click="">
<AppIcon
iconName="app-translate"
class="cursor color-secondary mr-16 ml-8"
style="font-size: 20px"
@click=""
>
</AppIcon>
</el-dropdown>
<Avatar></Avatar>
@ -41,15 +63,15 @@
import TopMenu from './top-menu/index.vue'
import Avatar from './avatar/index.vue'
import { useRouter } from 'vue-router'
import { langList } from '@/locales/index';
import { useLocale } from '@/locales/useLocale';
import { langList } from '@/locales/index'
import { useLocale } from '@/locales/useLocale'
const router = useRouter()
const defaultTitle = import.meta.env.VITE_APP_TITLE
const { changeLocale } = useLocale();
const { changeLocale } = useLocale()
const changeLang = (lang: string) => {
changeLocale(lang);
};
changeLocale(lang)
}
function toUrl(url: string) {
window.open(url, '_blank')
}

29
ui/src/views/template/component/CreateModelDialog.vue
View File

@ -81,11 +81,8 @@
</div>
<el-tooltip effect="dark" placement="right">
<template #content>
<p>为供应商的 LLM 模型，支持自定义输入</p>
<p>
下拉选项是 OpenAI
常用的一些大语言模型如：gpt-3.5-turbo-0613、gpt-3.5-turbo、gpt-4 等
</p>
<p>若下拉选项没有列出想要添加的LLM模型，自定义输入模型名称后回车即可</p>
<p>注意，基础模型需要与供应商的模型名称一致</p>
</template>
<AppIcon iconName="app-warning" class="app-warning-icon"></AppIcon>
</el-tooltip>
@ -101,12 +98,21 @@
allow-create
default-first-option
>
<el-option
v-for="item in base_model_list"
:key="item.name"
:label="item.name"
:value="item.name"
></el-option>
<el-option v-for="item in base_model_list" :key="item.name" :value="item.name">
<template #default>
<div class="flex align-center" style="display: inline-flex">
<div class="flex-between mr-4">
<span>{{ item.name }} </span>
</div>
<el-tooltip effect="dark" placement="right" v-if="item.desc">
<template #content>
<p>{{ item.desc }}</p>
</template>
<AppIcon iconName="app-warning" class="app-warning-icon"></AppIcon>
</el-tooltip>
</div>
</template>
</el-option>
</el-select>
</el-form-item>
</template>
@ -204,6 +210,7 @@ const list_base_model = (model_type: any) => {
)
}
}
const close = () => {
base_form_data.value = { name: '', model_type: '', model_name: '' }
credential_form_data.value = {}

28
ui/src/views/template/component/EditModel.vue
View File

@ -75,11 +75,8 @@
</div>
<el-tooltip effect="dark" placement="right">
<template #content>
<p>为供应商的 LLM 模型，支持自定义输入</p>
<p>
下拉选项是 OpenAI
常用的一些大语言模型如：gpt-3.5-turbo-0613、gpt-3.5-turbo、gpt-4 等
</p>
<p>若下拉选项没有列出想要添加的LLM模型，自定义输入模型名称后回车即可</p>
<p>注意，基础模型需要与供应商的模型名称一致</p>
</template>
<AppIcon iconName="app-warning" class="app-warning-icon"></AppIcon>
</el-tooltip>
@ -95,12 +92,21 @@
allow-create
default-first-option
>
<el-option
v-for="item in base_model_list"
:key="item.name"
:label="item.name"
:value="item.name"
></el-option>
<el-option v-for="item in base_model_list" :key="item.name" :value="item.name">
<template #default>
<div class="flex align-center" style="display: inline-flex">
<div class="flex-between mr-4">
<span>{{ item.name }} </span>
</div>
<el-tooltip effect="dark" placement="right" v-if="item.desc">
<template #content>
<p>{{ item.desc }}</p>
</template>
<AppIcon iconName="app-warning" class="app-warning-icon"></AppIcon>
</el-tooltip>
</div>
</template>
</el-option>
</el-select>
</el-form-item>
</template>