构建Agentic RAG AI Agent智能体教程 -- 知识铺

1
2

names = ['c-rag', 'self-rag', 'kg-rag']
files = ['../../data/c-rag.pdf', '../../data/self-rag.pdf', '../../data/kg-rag.pdf']

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43

# 采用chroma向量数据库
chroma = chromadb.HttpClient(host="localhost", port=8000)
collection = chroma.get_or_create_collection(name="agentic_rag")
vector_store = ChromaVectorStore(chroma_collection=collection)

# 创建针对某个文档的tool_agent
def create_tool_agent(file, name):
    # 文档拆分
    print(f'Starting to create tool agent for 【{name}】...\n')
    docs = SimpleDirectoryReader(input_files=[file]).load_data()
    splitter = SentenceSplitter(chunk_size=500, chunk_overlap=50)
    nodes = splitter.get_nodes_from_documents(docs)

    # 创建向量索引，并做持久保存
    if not os.path.exists(f"./storage/{name}"):
        print('Creating vector index...\n')
        storage_context = StorageContext.from_defaults(vector_store=vector_store)
        vector_index = VectorStoreIndex(nodes, storage_context=storage_context)
        vector_index.storage_context.persist(persist_dir=f"./storage/{name}")
    else:
        print('Loading vector index...\n')
        storage_context = StorageContext.from_defaults(persist_dir=f"./storage/{name}", vector_store=vector_store)
        vector_index = load_index_from_storage(storage_context=storage_context)

    # 创建基于向量的查询引擎
    query_engine = vector_index.as_query_engine(similarity_top_k=5)

    # 创建摘要索引与对应的查询引擎
    summary_index = SummaryIndex(nodes)
    summary_engine = summary_index.as_query_engine(response_mode="tree_summarize")

    # 将RAG引擎转化为两个tool
    query_tool = QueryEngineTool.from_defaults(query_engine=query_engine, name=f'query_tool', description=f'Use if you want to query details about {name}')
    summary_tool = QueryEngineTool.from_defaults(query_engine=summary_engine, name=f'summary_tool', description=f'Use ONLY IF you want to get a holistic summary of the documents. DO NOT USE if you want to query some details about {name}.')

    # 创建一个tool agent
    tool_agent = ReActAgent.from_tools([query_tool, summary_tool], verbose=True, system_prompt=f"""
    You are a specialized agent designed to answer queries about {name}.
    You must ALWAYS use at least one of the tools provided when answering a question; 
    DO NOT rely on prior knowledge. 
    DO NOT fabricate answer.
    """)
    return tool_agent

1
2
3
4
5
6
7
8

# 创建不同文档的agent
print('===============================================\n')
print('Creating tool agents for different documents...\n')
tool_agents_dict = {}

for name, file in zip(names, files):
    tool_agent = create_tool_agent(file, name)
    tool_agents_dict[name] = tool_agent

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28

# 首先将Tool Agent进行“工具化”
print('===============================================\n')
print('Creating tools from tool agents...\n')
all_tools = []

for name in names:
    agent_tool = QueryEngineTool.from_defaults(
        # 注意，Agent本身也是一种Query Engine，所以直接转为tool
        query_engine=tool_agents_dict[name],
        # 这个工具的名字
        name=f"tool_{name.replace("-", "")}",
        # 描述这个工具的作用和使用方法
        description=f"Use this tool if you want to answer any questions about {name}."
    )
    all_tools.append(agent_tool)

# 创建Top Agent
print('Creating top agent...\n')
top_agent = OpenAIAgent.from_tools(
    tools=all_tools,
    verbose=True,
    system_prompt="""
    You are an agent designed to answer queries over a set of given papers.
    Please always use the tools provided to answer a question.
    Do not rely on prior knowledge.
    DO NOT fabricate answer.
    """
)

1

top_agent.chat_repl()

1
2
3
4
5

# 创建工具检索器
print('===============================================\n')
print('Creating tool retrieve index...\n')
obj_index = ObjectIndex.from_objects(all_tools, index_cls=VectorStoreIndex)
tool_retriever = obj_index.as_retriever(similarity_top_k=5, verbose=True)

 1
 2
 3
 4
 5
 6
 7
 8
 9
10
11

# 创建顶级代理
top_agent = OpenAIAgent.from_tools(
    tools=all_tools,
    tool_retriever=tool_retriever,
    verbose=True,
    system_prompt="""
    You are an agent designed to answer queries over a set of given papers.
    Please always use the tools provided to answer a question.
    Do not rely on prior knowledge.
    """
)

1
2
3
4
5
6
7

# 检索需要的工具
tools_needed = tool_retriever.retrieve("What is the Adaptive retrieval in the c-RAG?")

# 打印需要的工具
print('Tools needed to answer the question:')
for tool in tools_needed:
    print(tool.metadata.name)