mirror of
https://github.com/csunny/DB-GPT.git
synced 2025-08-05 10:29:36 +00:00
167d972093
Co-authored-by: hzh97 <2976151305@qq.com> Co-authored-by: Fangyin Cheng <staneyffer@gmail.com> Co-authored-by: licunxing <864255598@qq.com>
63 lines
3.1 KiB
Python
63 lines
3.1 KiB
Python
from dbgpt._private.config import Config
|
||
from dbgpt.app.scene import AppScenePromptTemplateAdapter, ChatScene
|
||
from dbgpt.app.scene.chat_normal.out_parser import NormalChatOutputParser
|
||
from dbgpt.core import (
|
||
ChatPromptTemplate,
|
||
HumanPromptTemplate,
|
||
MessagesPlaceholder,
|
||
SystemPromptTemplate,
|
||
)
|
||
|
||
CFG = Config()
|
||
|
||
PROMPT_SCENE_DEFINE = """A chat between a curious user and an artificial intelligence assistant, who very familiar with database related knowledge.
|
||
The assistant gives helpful, detailed, professional and polite answers to the user's questions. """
|
||
|
||
|
||
_DEFAULT_TEMPLATE_ZH = """ 基于以下给出的已知信息, 准守规范约束，专业、简要回答用户的问题.
|
||
规范约束:
|
||
1.如果已知信息包含的图片、链接、表格、代码块等特殊markdown标签格式的信息，确保在答案中包含原文这些图片、链接、表格和代码标签，不要丢弃不要修改，如:图片格式：, 链接格式:[xxx](xxx), 表格格式:|xxx|xxx|xxx|, 代码格式:```xxx```.
|
||
2.如果无法从提供的内容中获取答案, 请说: "知识库中提供的内容不足以回答此问题" 禁止胡乱编造.
|
||
3.回答的时候最好按照1.2.3.点进行总结, 并以markdwon格式显示.
|
||
已知内容:
|
||
{context}
|
||
问题:
|
||
{question},请使用和用户相同的语言进行回答.
|
||
"""
|
||
_DEFAULT_TEMPLATE_EN = """ Based on the known information below, provide users with professional and concise answers to their questions.
|
||
constraints:
|
||
1.Ensure to include original markdown formatting elements such as images, links, tables, or code blocks without alteration in the response if they are present in the provided information.
|
||
For example, image format should be , link format [xxx](xxx), table format should be represented with |xxx|xxx|xxx|, and code format with xxx.
|
||
2.If the information available in the knowledge base is insufficient to answer the question, state clearly: "The content provided in the knowledge base is not enough to answer this question," and avoid making up answers.
|
||
3.When responding, it is best to summarize the points in the order of 1, 2, 3, And displayed in markdwon format.
|
||
known information:
|
||
{context}
|
||
question:
|
||
{question},when answering, use the same language as the "user".
|
||
"""
|
||
|
||
_DEFAULT_TEMPLATE = (
|
||
_DEFAULT_TEMPLATE_EN if CFG.LANGUAGE == "en" else _DEFAULT_TEMPLATE_ZH
|
||
)
|
||
|
||
PROMPT_NEED_STREAM_OUT = True
|
||
prompt = ChatPromptTemplate(
|
||
messages=[
|
||
SystemPromptTemplate.from_template(_DEFAULT_TEMPLATE),
|
||
MessagesPlaceholder(variable_name="chat_history"),
|
||
HumanPromptTemplate.from_template("{question}"),
|
||
]
|
||
)
|
||
|
||
prompt_adapter = AppScenePromptTemplateAdapter(
|
||
prompt=prompt,
|
||
template_scene=ChatScene.ChatKnowledge.value(),
|
||
stream_out=PROMPT_NEED_STREAM_OUT,
|
||
output_parser=NormalChatOutputParser(is_stream_out=PROMPT_NEED_STREAM_OUT),
|
||
need_historical_messages=False,
|
||
)
|
||
|
||
CFG.prompt_template_registry.register(
|
||
prompt_adapter, language=CFG.LANGUAGE, is_default=True
|
||
)
|