LlamaIndex 应用完整指南

pip install llama-index

import os
os.environ["OPENAI_API_KEY"] = "你的OpenAI API密钥"

from llama_index.core import VectorStoreIndex, SimpleDirectoryReader

# 加载文档
documents = SimpleDirectoryReader("data").load_data()

# 创建索引
index = VectorStoreIndex.from_documents(documents)

# 创建查询引擎
query_engine = index.as_query_engine()

# 进行查询
response = query_engine.query("你的问题")
print(response)

# 导入PDF文档
from llama_index.core import SimpleDirectoryReader

documents = SimpleDirectoryReader(
    input_dir="./data",
    filename_as_id=True
).load_data()

# 创建向量存储索引
from llama_index.core import VectorStoreIndex

index = VectorStoreIndex.from_documents(documents)

# 保存和加载索引
index.storage_context.persist()

# 创建查询引擎
query_engine = index.as_query_engine()

# 执行查询
response = query_engine.query("请总结这些文档的主要内容")

from llama_index.core.agent import ReActAgent
from llama_index.core.tools import QueryEngineTool

# 创建工具
tools = [
    QueryEngineTool(
        query_engine=query_engine,
        name="document_search",
        description="搜索文档内容"
    )
]

# 创建代理
agent = ReActAgent.from_tools(tools)

# 执行任务
response = agent.chat("分析文档中的关键信息")

# 创建聊天引擎
chat_engine = index.as_chat_engine(
    chat_mode="condense_question",  # 支持多种对话模式
    memory=ChatMemoryBuffer()       # 添加对话记忆
)

# 进行对话
response = chat_engine.chat("你好，请介绍一下文档的主要内容")

from llama_index.vector_stores import PineconeVectorStore

# 创建向量存储
vector_store = PineconeVectorStore(
    index_name="my-index",
    environment="us-west1-gcp"
)

# 使用向量存储创建索引
index = VectorStoreIndex.from_documents(
    documents,
    vector_store=vector_store
)

from llama_index.core.workflow import WorkflowGraph

# 创建工作流图
workflow = WorkflowGraph()

# 添加节点
workflow.add_node("loader", SimpleDirectoryReader("data"))
workflow.add_node("index", VectorStoreIndex)
workflow.add_node("query_engine", "index.as_query_engine()")

# 连接节点
workflow.connect("loader", "index")
workflow.connect("index", "query_engine")

# 运行工作流
results = workflow.run()

from llama_index.multi_modal_llms import OpenAIMultiModal

# 创建多模态LLM
llm = OpenAIMultiModal(model="gpt-4-vision-preview")

# 处理图像
response = llm.complete(
    prompt="描述这张图片",
    image_paths=["image.jpg"]
)

from llama_index.multi_modal_llms import OpenAIMultiModal

# 处理音频
response = llm.complete(
    prompt="转录这段音频",
    audio_paths=["audio.mp3"]
)

from llama_index.evaluation import QueryResponseEvaluator

# 创建评估器
evaluator = QueryResponseEvaluator()

# 评估响应
eval_result = evaluator.evaluate(
    query="问题",
    response="答案",
    reference_doc="参考文档"
)

from llama_index.cost_analysis import CostCalculator

# 创建成本计算器
calculator = CostCalculator()

# 计算成本
cost = calculator.calculate_cost(
    num_tokens=1000,
    model="gpt-4"
)