知识图谱操作（三）Neo4j-节点集自动递归控制查询

可视化问答系统搭载RAG增强检索

想要实现完整的RAG系统，基座技术是前几周实现的Faiss向量库和文本向量嵌入。

from transformers import GPT2Tokenizer, GPT2LMHeadModel

# 加载预训练的GPT模型和tokenizer
gpt_tokenizer = GPT2Tokenizer.from_pretrained('gpt2')
gpt_model = GPT2LMHeadModel.from_pretrained('gpt2')

def generate_answer(retrieved_docs, query):
    context = " ".join(retrieved_docs) + " " + query
    inputs = gpt_tokenizer.encode(context, return_tensors='pt', truncation=True, padding=True)
    outputs = gpt_model.generate(inputs, max_length=200, num_return_sequences=1)
    answer = gpt_tokenizer.decode(outputs[0], skip_special_tokens=True)
    return answer

# 生成回答
answer = generate_answer(retrieved_docs, query)
print("Generated Answer:", answer)

根据查询从Faiss数据库中索引相关知识

def retrieve(query, k=5):
    query_em