选择器示例

from langchain_community.embeddings import OllamaEmbeddings
from langchain_community.llms.ollama import Ollama
from langchain_community.vectorstores.faiss import FAISS
from langchain_core.example_selectors import SemanticSimilarityExampleSelector
from langchain_core.prompts import ChatPromptTemplate, PromptTemplate, FewShotPromptTemplate

llm = Ollama(model="qwen:7b")

template = '''
    示例输入:{input}, 示例输出:{output}
'''
example_prompt = PromptTemplate(
    input_variables=['input','output'],
    template=template
)

examples = [
    {"input":"海盗", "output":"船"},
    {"input":"飞行员", "output":"飞机"},
    {"input":"驾驶员", "output":"车"},
    {"input":"树", "output":"地面"},
    {"input":"鸟", "output":"鸟巢"}
]

#调用示例选择器
#SemanticSimilarityExampleSelector 将根据语义选择与您的输入相似的示例
example_selector = SemanticSimilarityExampleSelector.from_examples(
    #可供选择的示例模板
    examples,
    #测量语义相似性的嵌入的嵌入类
    OllamaEmbeddings(),
    #存储嵌入和进行相似搜索的Vectortore类
    FAISS,
    #要生成的示例数
    k=2
)


similar_propt = FewShotPromptTemplate(
    #有助于选择示例的对象
    example_selector = example_selector,
    #提示词
    example_prompt = example_prompt,
    #将添加到提示顶部和底部的自定义项
    prefix = "根据下面示例，写出输出",
    suffix = "输入:{noun}\n输出:",
    #你的提示词接受的输入
    input_variables=['noun']
)

final_prompt = similar_propt.format(noun='硬盘')
print()

response = llm(final_prompt)

print(response)