从零搭建RAG知识库：基于腾讯云Embedding模型完整部署教程

pip install tencentcloud-sdk-python chromadb langchain python-dotenv

TENCENT_SECRET_ID=你的腾讯云SecretId
TENCENT_SECRET_KEY=你的腾讯云SecretKey
TENCENT_REGION=ap-guangzhou

import os
from dotenv import load_dotenv
from tencentcloud.common import credential
from tencentcloud.common.profile.client_profile import ClientProfile
from tencentcloud.common.profile.http_profile import HttpProfile
from tencentcloud.lke.v20231130 import lke_client, models

# 加载环境变量
load_dotenv()

class TencentEmbedding:
    def __init__(self):
        # 初始化腾讯云认证
        cred = credential.Credential(
            os.getenv("TENCENT_SECRET_ID"),
            os.getenv("TENCENT_SECRET_KEY")
        )
        http_profile = HttpProfile()
        http_profile.endpoint = "lke.tencentcloudapi.com"
        client_profile = ClientProfile()
        client_profile.httpProfile = http_profile
        self.client = lke_client.LkeClient(cred, os.getenv("TENCENT_REGION"), client_profile)

    # 单文本向量化
    def embed_query(self, text: str):
        req = models.GetEmbeddingRequest()
        params = {
            "Text": text,
            "Model": "lke-text-embedding-v1"
        }
        req.from_json_string(str(params))
        resp = self.client.GetEmbedding(req)
        return resp.Embedding

    # 批量文档向量化
    def embed_documents(self, texts: list):
        embeddings = []
        for text in texts:
            embeddings.append(self.embed_query(text))
        return embeddings

from langchain.text_splitter import RecursiveCharacterTextSplitter
from langchain.document_loaders import TextLoader
from langchain.vectorstores import Chroma

# 初始化自定义腾讯Embedding
embedding = TencentEmbedding()

# 1. 加载本地文档
loader = TextLoader("test_doc.txt", encoding="utf-8")
documents = loader.load()

# 2. 文本切片（适配中文场景）
text_splitter = RecursiveCharacterTextSplitter(
    chunk_size=800,
    chunk_overlap=150,
    separators=["\n\n", "\n", "。", "，"]
)
split_docs = text_splitter.split_documents(documents)

# 3. 腾讯Embedding向量化 + 向量入库
vectordb = Chroma.from_documents(
    documents=split_docs,
    embedding=embedding,
    persist_directory="./tencent_rag_db"
)
# 持久化向量库
vectordb.persist()
print("文档向量入库完成！")

from langchain.chains import RetrievalQA
from langchain.llms import OpenAI
from langchain.prompts import PromptTemplate

# 自定义RAG提示词（约束模型基于知识库回答）
prompt_template = """
请严格基于以下知识库内容回答用户问题，禁止编造信息，无匹配内容则如实告知。
知识库内容：{context}
用户问题：{question}
"""
prompt = PromptTemplate(template=prompt_template, input_variables=["context", "question"])

# 加载持久化向量库
vectordb = Chroma(
    persist_directory="./tencent_rag_db",
    embedding_function=embedding
)

# 构建检索器：召回Top3相似度最高文档片段
retriever = vectordb.as_retriever(search_kwargs={"k": 3})

# 搭建RAG问答链
qa_chain = RetrievalQA.from_chain_type(
    llm=OpenAI(temperature=0),
    chain_type="stuff",
    retriever=retriever,
    chain_type_kwargs={"prompt": prompt}
)

# 测试问答
if __name__ == "__main__":
    res = qa_chain.run("你的自定义问题")
    print("RAG回答结果：", res)