大家好，这里是程序员晚枫，正在all in AI编程实战。

第15讲：RAG入门——让AI读取你的文档

什么是RAG？

RAG = Retrieval-Augmented Generation（检索增强生成）

简单说：先从你的文档里找到相关内容，再让AI基于这些内容回答问题。

为什么需要RAG？因为AI模型的知识有截止日期，而且不知道你公司的内部文档。RAG让AI"读过"你的资料再回答。

1、RAG的工作流程

1	用户提问 → 检索相关文档 → 把文档+问题发给AI → AI基于文档回答

2、简单实现：直接把文档塞给AI

文档不大时，最简单的方式：

from openai import OpenAI

client = OpenAI(api_key="你的Key", base_url="https://api.deepseek.com")

with open('产品说明书.txt', 'r', encoding='utf-8') as f:
    doc_content = f.read()

question = "这个产品的保修期是多久？"

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[
        {"role": "system", "content": f"请根据以下文档内容回答用户问题。如果文档中没有相关信息，请说'文档中未找到相关信息'。\n\n文档内容：\n{doc_content}"},
        {"role": "user", "content": question}
    ]
)

print(response.choices[0].message.content)

3、处理大文档：分段检索

文档太大怎么办？分成小段，只检索相关的：

def split_text(text, chunk_size=500):
    chunks = []
    for i in range(0, len(text), chunk_size):
        chunks.append(text[i:i+chunk_size])
    return chunks

def simple_search(chunks, keyword):
    results = []
    for i, chunk in enumerate(chunks):
        if keyword in chunk:
            results.append((i, chunk))
    return results

with open('大文档.txt', 'r', encoding='utf-8') as f:
    text = f.read()

chunks = split_text(text)
question = "退款流程是什么？"
results = simple_search(chunks, "退款")

context = "\n".join([chunk for _, chunk in results[:3]])

response = client.chat.completions.create(
    model="deepseek-chat",
    messages=[
        {"role": "system", "content": f"根据以下内容回答：\n{context}"},
        {"role": "user", "content": question}
    ]
)

4、用向量检索提升效果

关键词搜索太简单，用向量搜索更智能：

1	pip install chromadb sentence-transformers

import chromadb
from sentence_transformers import SentenceTransformer

model = SentenceTransformer('shibing624/text2vec-base-chinese')
client_db = chromadb.Client()
collection = client_db.create_collection("my_docs")

with open('文档.txt', 'r', encoding='utf-8') as f:
    text = f.read()

chunks = split_text(text, 500)

for i, chunk in enumerate(chunks):
    collection.add(documents=[chunk], ids=[f"chunk_{i}"])

results = collection.query(query_texts=["退款流程是什么？"], n_results=3)
context = "\n".join(results['documents'][0])

5、实战：做一个文档问答助手

from openai import OpenAI
import chromadb

ai_client = OpenAI(api_key="你的Key", base_url="https://api.deepseek.com")
db_client = chromadb.Client()
collection = db_client.get_or_create_collection("docs")

def add_document(file_path):
    with open(file_path, 'r', encoding='utf-8') as f:
        text = f.read()
    chunks = split_text(text, 500)
    for i, chunk in enumerate(chunks):
        collection.add(documents=[chunk], ids=[f"{file_path}_{i}"])
    print(f"已添加 {len(chunks)} 个文档片段")

def ask(question):
    results = collection.query(query_texts=[question], n_results=3)
    context = "\n".join(results['documents'][0])
    response = ai_client.chat.completions.create(
        model="deepseek-chat",
        messages=[
            {"role": "system", "content": f"根据以下内容回答问题：\n{context}"},
            {"role": "user", "content": question}
        ]
    )
    return response.choices[0].message.content

add_document('产品手册.txt')
add_document('常见问题.txt')
print(ask("如何退换货？"))